Created
May 12, 2020 23:22
-
-
Save benvanik/9a94c3a4b23d4ebbd4e4c23ccdd20377 to your computer and use it in GitHub Desktop.
resnet50 IR
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 370 : i32}} { | |
| flow.variable @"__iree_flow___sm_node186__m.layer-2.kernel" dense<1.200000e+00> : tensor<7x7x3x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node187__m.layer-2.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node193__m.layer-3.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node194__m.layer-3.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node195__m.layer-3.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node196__m.layer-3.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node213__m.layer-7.kernel" dense<1.200000e+00> : tensor<1x1x64x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node214__m.layer-7.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node220__m.layer-8.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node221__m.layer-8.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node222__m.layer-8.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node223__m.layer-8.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node232__m.layer-10.kernel" dense<1.200000e+00> : tensor<3x3x64x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node233__m.layer-10.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node239__m.layer-11.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node240__m.layer-11.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node241__m.layer-11.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node242__m.layer-11.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node251__m.layer-13.kernel" dense<1.200000e+00> : tensor<1x1x64x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node252__m.layer-13.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node257__m.layer-14.kernel" dense<1.200000e+00> : tensor<1x1x64x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node258__m.layer-14.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node264__m.layer-15.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node265__m.layer-15.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node266__m.layer-15.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node267__m.layer-15.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node273__m.layer-16.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node274__m.layer-16.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node275__m.layer-16.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node276__m.layer-16.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node289__m.layer-19.kernel" dense<1.200000e+00> : tensor<1x1x256x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node290__m.layer-19.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node296__m.layer-20.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node297__m.layer-20.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node298__m.layer-20.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node299__m.layer-20.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node308__m.layer-22.kernel" dense<1.200000e+00> : tensor<3x3x64x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node309__m.layer-22.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node315__m.layer-23.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node316__m.layer-23.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node317__m.layer-23.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node318__m.layer-23.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node327__m.layer-25.kernel" dense<1.200000e+00> : tensor<1x1x64x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node328__m.layer-25.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node334__m.layer-26.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node335__m.layer-26.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node336__m.layer-26.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node337__m.layer-26.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node350__m.layer-29.kernel" dense<1.200000e+00> : tensor<1x1x256x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node351__m.layer-29.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node357__m.layer-30.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node358__m.layer-30.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node359__m.layer-30.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node360__m.layer-30.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node369__m.layer-32.kernel" dense<1.200000e+00> : tensor<3x3x64x64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node370__m.layer-32.bias" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node376__m.layer-33.gamma" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node377__m.layer-33.beta" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node378__m.layer-33.moving_mean" dense<0.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node379__m.layer-33.moving_variance" dense<1.000000e+00> : tensor<64xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node388__m.layer-35.kernel" dense<1.200000e+00> : tensor<1x1x64x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node389__m.layer-35.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node395__m.layer-36.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node396__m.layer-36.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node397__m.layer-36.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node398__m.layer-36.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node411__m.layer-39.kernel" dense<1.200000e+00> : tensor<1x1x256x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node412__m.layer-39.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node418__m.layer-40.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node419__m.layer-40.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node420__m.layer-40.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node421__m.layer-40.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node430__m.layer-42.kernel" dense<1.200000e+00> : tensor<3x3x128x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node431__m.layer-42.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node437__m.layer-43.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node438__m.layer-43.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node439__m.layer-43.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node440__m.layer-43.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node449__m.layer-45.kernel" dense<1.200000e+00> : tensor<1x1x256x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node450__m.layer-45.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node455__m.layer-46.kernel" dense<1.200000e+00> : tensor<1x1x128x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node456__m.layer-46.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node462__m.layer-47.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node463__m.layer-47.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node464__m.layer-47.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node465__m.layer-47.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node471__m.layer-48.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node472__m.layer-48.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node473__m.layer-48.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node474__m.layer-48.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node487__m.layer-51.kernel" dense<1.200000e+00> : tensor<1x1x512x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node488__m.layer-51.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node494__m.layer-52.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node495__m.layer-52.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node496__m.layer-52.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node497__m.layer-52.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node506__m.layer-54.kernel" dense<1.200000e+00> : tensor<3x3x128x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node507__m.layer-54.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node513__m.layer-55.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node514__m.layer-55.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node515__m.layer-55.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node516__m.layer-55.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node525__m.layer-57.kernel" dense<1.200000e+00> : tensor<1x1x128x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node526__m.layer-57.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node532__m.layer-58.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node533__m.layer-58.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node534__m.layer-58.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node535__m.layer-58.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node548__m.layer-61.kernel" dense<1.200000e+00> : tensor<1x1x512x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node549__m.layer-61.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node555__m.layer-62.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node556__m.layer-62.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node557__m.layer-62.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node558__m.layer-62.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node567__m.layer-64.kernel" dense<1.200000e+00> : tensor<3x3x128x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node568__m.layer-64.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node574__m.layer-65.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node575__m.layer-65.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node576__m.layer-65.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node577__m.layer-65.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node586__m.layer-67.kernel" dense<1.200000e+00> : tensor<1x1x128x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node587__m.layer-67.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node593__m.layer-68.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node594__m.layer-68.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node595__m.layer-68.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node596__m.layer-68.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node609__m.layer-71.kernel" dense<1.200000e+00> : tensor<1x1x512x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node610__m.layer-71.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node616__m.layer-72.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node617__m.layer-72.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node618__m.layer-72.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node619__m.layer-72.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node628__m.layer-74.kernel" dense<1.200000e+00> : tensor<3x3x128x128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node629__m.layer-74.bias" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node635__m.layer-75.gamma" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node636__m.layer-75.beta" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node637__m.layer-75.moving_mean" dense<0.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node638__m.layer-75.moving_variance" dense<1.000000e+00> : tensor<128xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node647__m.layer-77.kernel" dense<1.200000e+00> : tensor<1x1x128x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node648__m.layer-77.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node654__m.layer-78.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node655__m.layer-78.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node656__m.layer-78.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node657__m.layer-78.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node670__m.layer-81.kernel" dense<1.200000e+00> : tensor<1x1x512x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node671__m.layer-81.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node677__m.layer-82.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node678__m.layer-82.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node679__m.layer-82.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node680__m.layer-82.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node689__m.layer-84.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node690__m.layer-84.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node696__m.layer-85.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node697__m.layer-85.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node698__m.layer-85.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node699__m.layer-85.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node708__m.layer-87.kernel" dense<1.200000e+00> : tensor<1x1x512x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node709__m.layer-87.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node714__m.layer-88.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node715__m.layer-88.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node721__m.layer-89.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node722__m.layer-89.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node723__m.layer-89.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node724__m.layer-89.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node730__m.layer-90.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node731__m.layer-90.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node732__m.layer-90.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node733__m.layer-90.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node746__m.layer-93.kernel" dense<1.200000e+00> : tensor<1x1x1024x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node747__m.layer-93.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node753__m.layer-94.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node754__m.layer-94.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node755__m.layer-94.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node756__m.layer-94.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node765__m.layer-96.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node766__m.layer-96.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node772__m.layer-97.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node773__m.layer-97.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node774__m.layer-97.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node775__m.layer-97.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node784__m.layer-99.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node785__m.layer-99.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node791__m.layer-100.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node792__m.layer-100.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node793__m.layer-100.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node794__m.layer-100.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node807__m.layer-103.kernel" dense<1.200000e+00> : tensor<1x1x1024x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node808__m.layer-103.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node814__m.layer-104.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node815__m.layer-104.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node816__m.layer-104.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node817__m.layer-104.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node826__m.layer-106.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node827__m.layer-106.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node833__m.layer-107.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node834__m.layer-107.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node835__m.layer-107.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node836__m.layer-107.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node845__m.layer-109.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node846__m.layer-109.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node852__m.layer-110.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node853__m.layer-110.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node854__m.layer-110.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node855__m.layer-110.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node868__m.layer-113.kernel" dense<1.200000e+00> : tensor<1x1x1024x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node869__m.layer-113.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node875__m.layer-114.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node876__m.layer-114.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node877__m.layer-114.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node878__m.layer-114.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node887__m.layer-116.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node888__m.layer-116.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node894__m.layer-117.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node895__m.layer-117.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node896__m.layer-117.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node897__m.layer-117.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node906__m.layer-119.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node907__m.layer-119.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node913__m.layer-120.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node914__m.layer-120.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node915__m.layer-120.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node916__m.layer-120.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node929__m.layer-123.kernel" dense<1.200000e+00> : tensor<1x1x1024x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node930__m.layer-123.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node936__m.layer-124.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node937__m.layer-124.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node938__m.layer-124.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node939__m.layer-124.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node948__m.layer-126.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node949__m.layer-126.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node955__m.layer-127.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node956__m.layer-127.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node957__m.layer-127.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node958__m.layer-127.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node967__m.layer-129.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node968__m.layer-129.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node974__m.layer-130.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node975__m.layer-130.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node976__m.layer-130.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node977__m.layer-130.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node990__m.layer-133.kernel" dense<1.200000e+00> : tensor<1x1x1024x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node991__m.layer-133.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node997__m.layer-134.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node998__m.layer-134.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node999__m.layer-134.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1000__m.layer-134.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1009__m.layer-136.kernel" dense<1.200000e+00> : tensor<3x3x256x256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1010__m.layer-136.bias" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1016__m.layer-137.gamma" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1017__m.layer-137.beta" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1018__m.layer-137.moving_mean" dense<0.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1019__m.layer-137.moving_variance" dense<1.000000e+00> : tensor<256xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1028__m.layer-139.kernel" dense<1.200000e+00> : tensor<1x1x256x1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1029__m.layer-139.bias" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1035__m.layer-140.gamma" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1036__m.layer-140.beta" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1037__m.layer-140.moving_mean" dense<0.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1038__m.layer-140.moving_variance" dense<1.000000e+00> : tensor<1024xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1051__m.layer-143.kernel" dense<1.200000e+00> : tensor<1x1x1024x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1052__m.layer-143.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1058__m.layer-144.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1059__m.layer-144.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1060__m.layer-144.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1061__m.layer-144.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1070__m.layer-146.kernel" dense<1.200000e+00> : tensor<3x3x512x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1071__m.layer-146.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1077__m.layer-147.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1078__m.layer-147.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1079__m.layer-147.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1080__m.layer-147.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1089__m.layer-149.kernel" dense<1.200000e+00> : tensor<1x1x1024x2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1090__m.layer-149.bias" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1095__m.layer-150.kernel" dense<1.200000e+00> : tensor<1x1x512x2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1096__m.layer-150.bias" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1102__m.layer-151.gamma" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1103__m.layer-151.beta" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1104__m.layer-151.moving_mean" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1105__m.layer-151.moving_variance" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1111__m.layer-152.gamma" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1112__m.layer-152.beta" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1113__m.layer-152.moving_mean" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1114__m.layer-152.moving_variance" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1127__m.layer-155.kernel" dense<1.200000e+00> : tensor<1x1x2048x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1128__m.layer-155.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1134__m.layer-156.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1135__m.layer-156.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1136__m.layer-156.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1137__m.layer-156.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1146__m.layer-158.kernel" dense<1.200000e+00> : tensor<3x3x512x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1147__m.layer-158.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1153__m.layer-159.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1154__m.layer-159.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1155__m.layer-159.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1156__m.layer-159.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1165__m.layer-161.kernel" dense<1.200000e+00> : tensor<1x1x512x2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1166__m.layer-161.bias" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1172__m.layer-162.gamma" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1173__m.layer-162.beta" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1174__m.layer-162.moving_mean" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1175__m.layer-162.moving_variance" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1188__m.layer-165.kernel" dense<1.200000e+00> : tensor<1x1x2048x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1189__m.layer-165.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1195__m.layer-166.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1196__m.layer-166.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1197__m.layer-166.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1198__m.layer-166.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1207__m.layer-168.kernel" dense<1.200000e+00> : tensor<3x3x512x512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1208__m.layer-168.bias" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1214__m.layer-169.gamma" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1215__m.layer-169.beta" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1216__m.layer-169.moving_mean" dense<0.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1217__m.layer-169.moving_variance" dense<1.000000e+00> : tensor<512xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1226__m.layer-171.kernel" dense<1.200000e+00> : tensor<1x1x512x2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1227__m.layer-171.bias" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1233__m.layer-172.gamma" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1234__m.layer-172.beta" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1235__m.layer-172.moving_mean" dense<0.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| flow.variable @"__iree_flow___sm_node1236__m.layer-172.moving_variance" dense<1.000000e+00> : tensor<2048xf32> attributes {sym_visibility = "private"} | |
| func @predict() -> tensor<1x1x1x2048xf32> attributes {iree.module.export, iree.reflection = {abi = "sip", abiv = 1 : i32, sip = "I8!S5!k0_0R3!_0"}, tf._input_shapes = ["tfshape$dim { size: 1 } dim { size: 32 } dim { size: 32 } dim { size: 3 }", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true"], tf.signature.is_stateful} { | |
| %cst = constant dense<1.200000e+00> : tensor<1x32x32x3xf32> | |
| %0 = xla_hlo.constant dense<0xFF800000> : tensor<f32> | |
| %1 = xla_hlo.constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.200000e+00> : tensor<7x7x3x64xf32> | |
| %cst_1 = constant dense<1.200000e+00> : tensor<1x1x64x64xf32> | |
| %cst_2 = constant dense<1.200000e+00> : tensor<1x1x256x64xf32> | |
| %cst_3 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_4 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_5 = constant dense<1.200000e+00> : tensor<3x3x64x64xf32> | |
| %cst_6 = constant dense<1.200000e+00> : tensor<1x1x64x256xf32> | |
| %cst_7 = constant dense<1.200000e+00> : tensor<1x1x256x512xf32> | |
| %cst_8 = constant dense<1.200000e+00> : tensor<1x1x256x128xf32> | |
| %cst_9 = constant dense<1.200000e+00> : tensor<1x1x512x128xf32> | |
| %cst_10 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_11 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_12 = constant dense<1.200000e+00> : tensor<3x3x128x128xf32> | |
| %cst_13 = constant dense<1.200000e+00> : tensor<1x1x128x512xf32> | |
| %cst_14 = constant dense<1.200000e+00> : tensor<1x1x512x1024xf32> | |
| %cst_15 = constant dense<1.200000e+00> : tensor<1x1x512x256xf32> | |
| %cst_16 = constant dense<1.200000e+00> : tensor<1x1x1024x256xf32> | |
| %cst_17 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_18 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_19 = constant dense<1.200000e+00> : tensor<3x3x256x256xf32> | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.200000e+00> : tensor<1x1x256x1024xf32> | |
| %cst_23 = constant dense<1.200000e+00> : tensor<1x1x1024x2048xf32> | |
| %cst_24 = constant dense<1.200000e+00> : tensor<1x1x1024x512xf32> | |
| %cst_25 = constant dense<1.200000e+00> : tensor<1x1x2048x512xf32> | |
| %cst_26 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_27 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_28 = constant dense<1.200000e+00> : tensor<3x3x512x512xf32> | |
| %cst_29 = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_30 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_31 = constant dense<1.200000e+00> : tensor<1x1x512x2048xf32> | |
| %2 = iree.do_not_optimize(%cst) : tensor<1x32x32x3xf32> | |
| %3 = "xla_hlo.convolution"(%2, %cst_0) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<3> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x32x32x3xf32>, tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> | |
| %4 = "xla_hlo.add"(%3, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x16x16x64xf32>, tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %5 = "xla_hlo.batch_norm_inference"(%4, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x16x16x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %6 = "xla_hlo.maximum"(%5, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x16x16x64xf32>, tensor<f32>) -> tensor<1x16x16x64xf32> | |
| %7 = "xla_hlo.pad"(%6, %1) {edge_padding_high = dense<[0, 1, 1, 0]> : tensor<4xi64>, edge_padding_low = dense<[0, 1, 1, 0]> : tensor<4xi64>, interior_padding = dense<0> : tensor<4xi64>} : (tensor<1x16x16x64xf32>, tensor<f32>) -> tensor<1x18x18x64xf32> | |
| %8 = "xla_hlo.reduce_window"(%7, %0) ( { | |
| ^bb0(%arg0: tensor<f32>, %arg1: tensor<f32>): // no predecessors | |
| %229 = xla_hlo.maximum %arg0, %arg1 : tensor<f32> | |
| "xla_hlo.return"(%229) : (tensor<f32>) -> () | |
| }) {window_dimensions = dense<[1, 3, 3, 1]> : tensor<4xi64>, window_strides = dense<[1, 2, 2, 1]> : tensor<4xi64>} : (tensor<1x18x18x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %9 = "xla_hlo.convolution"(%8, %cst_6) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %10 = "xla_hlo.add"(%9, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %11 = "xla_hlo.batch_norm_inference"(%10, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %12 = "xla_hlo.convolution"(%8, %cst_1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %13 = "xla_hlo.add"(%12, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %14 = "xla_hlo.batch_norm_inference"(%13, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %15 = "xla_hlo.maximum"(%14, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %16 = "xla_hlo.convolution"(%15, %cst_5) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %17 = "xla_hlo.add"(%16, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %18 = "xla_hlo.batch_norm_inference"(%17, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %19 = "xla_hlo.maximum"(%18, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %20 = "xla_hlo.convolution"(%19, %cst_6) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %21 = "xla_hlo.add"(%20, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %22 = "xla_hlo.batch_norm_inference"(%21, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %23 = xla_hlo.add %11, %22 : tensor<1x8x8x256xf32> | |
| %24 = "xla_hlo.maximum"(%23, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x256xf32>, tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %25 = "xla_hlo.convolution"(%24, %cst_2) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| %26 = "xla_hlo.add"(%25, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %27 = "xla_hlo.batch_norm_inference"(%26, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %28 = "xla_hlo.maximum"(%27, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %29 = "xla_hlo.convolution"(%28, %cst_5) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %30 = "xla_hlo.add"(%29, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %31 = "xla_hlo.batch_norm_inference"(%30, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %32 = "xla_hlo.maximum"(%31, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %33 = "xla_hlo.convolution"(%32, %cst_6) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %34 = "xla_hlo.add"(%33, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %35 = "xla_hlo.batch_norm_inference"(%34, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %36 = xla_hlo.add %24, %35 : tensor<1x8x8x256xf32> | |
| %37 = "xla_hlo.maximum"(%36, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x256xf32>, tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %38 = "xla_hlo.convolution"(%37, %cst_2) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| %39 = "xla_hlo.add"(%38, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %40 = "xla_hlo.batch_norm_inference"(%39, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %41 = "xla_hlo.maximum"(%40, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %42 = "xla_hlo.convolution"(%41, %cst_5) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %43 = "xla_hlo.add"(%42, %cst_4) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %44 = "xla_hlo.batch_norm_inference"(%43, %cst_3, %cst_4, %cst_4, %cst_3) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>, tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %45 = "xla_hlo.maximum"(%44, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %46 = "xla_hlo.convolution"(%45, %cst_6) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %47 = "xla_hlo.add"(%46, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x8x8x256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %48 = "xla_hlo.batch_norm_inference"(%47, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x8x8x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %49 = xla_hlo.add %37, %48 : tensor<1x8x8x256xf32> | |
| %50 = "xla_hlo.maximum"(%49, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x8x8x256xf32>, tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %51 = "xla_hlo.convolution"(%50, %cst_7) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> | |
| %52 = "xla_hlo.add"(%51, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %53 = "xla_hlo.batch_norm_inference"(%52, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %54 = "xla_hlo.convolution"(%50, %cst_8) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> | |
| %55 = "xla_hlo.add"(%54, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %56 = "xla_hlo.batch_norm_inference"(%55, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %57 = "xla_hlo.maximum"(%56, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %58 = "xla_hlo.convolution"(%57, %cst_12) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %59 = "xla_hlo.add"(%58, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %60 = "xla_hlo.batch_norm_inference"(%59, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %61 = "xla_hlo.maximum"(%60, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %62 = "xla_hlo.convolution"(%61, %cst_13) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %63 = "xla_hlo.add"(%62, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %64 = "xla_hlo.batch_norm_inference"(%63, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %65 = xla_hlo.add %53, %64 : tensor<1x4x4x512xf32> | |
| %66 = "xla_hlo.maximum"(%65, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x512xf32>, tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %67 = "xla_hlo.convolution"(%66, %cst_9) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %68 = "xla_hlo.add"(%67, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %69 = "xla_hlo.batch_norm_inference"(%68, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %70 = "xla_hlo.maximum"(%69, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %71 = "xla_hlo.convolution"(%70, %cst_12) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %72 = "xla_hlo.add"(%71, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %73 = "xla_hlo.batch_norm_inference"(%72, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %74 = "xla_hlo.maximum"(%73, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %75 = "xla_hlo.convolution"(%74, %cst_13) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %76 = "xla_hlo.add"(%75, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %77 = "xla_hlo.batch_norm_inference"(%76, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %78 = xla_hlo.add %66, %77 : tensor<1x4x4x512xf32> | |
| %79 = "xla_hlo.maximum"(%78, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x512xf32>, tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %80 = "xla_hlo.convolution"(%79, %cst_9) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %81 = "xla_hlo.add"(%80, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %82 = "xla_hlo.batch_norm_inference"(%81, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %83 = "xla_hlo.maximum"(%82, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %84 = "xla_hlo.convolution"(%83, %cst_12) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %85 = "xla_hlo.add"(%84, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %86 = "xla_hlo.batch_norm_inference"(%85, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %87 = "xla_hlo.maximum"(%86, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %88 = "xla_hlo.convolution"(%87, %cst_13) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %89 = "xla_hlo.add"(%88, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %90 = "xla_hlo.batch_norm_inference"(%89, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %91 = xla_hlo.add %79, %90 : tensor<1x4x4x512xf32> | |
| %92 = "xla_hlo.maximum"(%91, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x512xf32>, tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %93 = "xla_hlo.convolution"(%92, %cst_9) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %94 = "xla_hlo.add"(%93, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %95 = "xla_hlo.batch_norm_inference"(%94, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %96 = "xla_hlo.maximum"(%95, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %97 = "xla_hlo.convolution"(%96, %cst_12) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %98 = "xla_hlo.add"(%97, %cst_11) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %99 = "xla_hlo.batch_norm_inference"(%98, %cst_10, %cst_11, %cst_11, %cst_10) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>, tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %100 = "xla_hlo.maximum"(%99, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x128xf32>, tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %101 = "xla_hlo.convolution"(%100, %cst_13) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %102 = "xla_hlo.add"(%101, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x4x4x512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %103 = "xla_hlo.batch_norm_inference"(%102, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x4x4x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %104 = xla_hlo.add %92, %103 : tensor<1x4x4x512xf32> | |
| %105 = "xla_hlo.maximum"(%104, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x4x4x512xf32>, tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %106 = "xla_hlo.convolution"(%105, %cst_14) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %107 = "xla_hlo.add"(%106, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.batch_norm_inference"(%107, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = "xla_hlo.convolution"(%105, %cst_15) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.add"(%109, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %111 = "xla_hlo.batch_norm_inference"(%110, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %112 = "xla_hlo.maximum"(%111, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %113 = "xla_hlo.convolution"(%112, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = "xla_hlo.add"(%113, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %115 = "xla_hlo.batch_norm_inference"(%114, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.maximum"(%115, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = "xla_hlo.convolution"(%116, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %118 = "xla_hlo.add"(%117, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %119 = "xla_hlo.batch_norm_inference"(%118, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %120 = xla_hlo.add %108, %119 : tensor<1x2x2x1024xf32> | |
| %121 = "xla_hlo.maximum"(%120, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %122 = "xla_hlo.convolution"(%121, %cst_16) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %123 = "xla_hlo.add"(%122, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %124 = "xla_hlo.batch_norm_inference"(%123, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %125 = "xla_hlo.maximum"(%124, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %126 = "xla_hlo.convolution"(%125, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %127 = "xla_hlo.add"(%126, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %128 = "xla_hlo.batch_norm_inference"(%127, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %129 = "xla_hlo.maximum"(%128, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %130 = "xla_hlo.convolution"(%129, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %131 = "xla_hlo.add"(%130, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %132 = "xla_hlo.batch_norm_inference"(%131, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %133 = xla_hlo.add %121, %132 : tensor<1x2x2x1024xf32> | |
| %134 = "xla_hlo.maximum"(%133, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %135 = "xla_hlo.convolution"(%134, %cst_16) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %136 = "xla_hlo.add"(%135, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %137 = "xla_hlo.batch_norm_inference"(%136, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %138 = "xla_hlo.maximum"(%137, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %139 = "xla_hlo.convolution"(%138, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %140 = "xla_hlo.add"(%139, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %141 = "xla_hlo.batch_norm_inference"(%140, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %142 = "xla_hlo.maximum"(%141, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %143 = "xla_hlo.convolution"(%142, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %144 = "xla_hlo.add"(%143, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %145 = "xla_hlo.batch_norm_inference"(%144, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %146 = xla_hlo.add %134, %145 : tensor<1x2x2x1024xf32> | |
| %147 = "xla_hlo.maximum"(%146, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %148 = "xla_hlo.convolution"(%147, %cst_16) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %149 = "xla_hlo.add"(%148, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %150 = "xla_hlo.batch_norm_inference"(%149, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %151 = "xla_hlo.maximum"(%150, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %152 = "xla_hlo.convolution"(%151, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %153 = "xla_hlo.add"(%152, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %154 = "xla_hlo.batch_norm_inference"(%153, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %155 = "xla_hlo.maximum"(%154, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %156 = "xla_hlo.convolution"(%155, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %157 = "xla_hlo.add"(%156, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %158 = "xla_hlo.batch_norm_inference"(%157, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %159 = xla_hlo.add %147, %158 : tensor<1x2x2x1024xf32> | |
| %160 = "xla_hlo.maximum"(%159, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %161 = "xla_hlo.convolution"(%160, %cst_16) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %162 = "xla_hlo.add"(%161, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %163 = "xla_hlo.batch_norm_inference"(%162, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %164 = "xla_hlo.maximum"(%163, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %165 = "xla_hlo.convolution"(%164, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %166 = "xla_hlo.add"(%165, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %167 = "xla_hlo.batch_norm_inference"(%166, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %168 = "xla_hlo.maximum"(%167, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %169 = "xla_hlo.convolution"(%168, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %170 = "xla_hlo.add"(%169, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %171 = "xla_hlo.batch_norm_inference"(%170, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %172 = xla_hlo.add %160, %171 : tensor<1x2x2x1024xf32> | |
| %173 = "xla_hlo.maximum"(%172, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %174 = "xla_hlo.convolution"(%173, %cst_16) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %175 = "xla_hlo.add"(%174, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %176 = "xla_hlo.batch_norm_inference"(%175, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %177 = "xla_hlo.maximum"(%176, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %178 = "xla_hlo.convolution"(%177, %cst_19) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %179 = "xla_hlo.add"(%178, %cst_18) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %180 = "xla_hlo.batch_norm_inference"(%179, %cst_17, %cst_18, %cst_18, %cst_17) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>, tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %181 = "xla_hlo.maximum"(%180, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x256xf32>, tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %182 = "xla_hlo.convolution"(%181, %cst_22) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %183 = "xla_hlo.add"(%182, %cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %184 = "xla_hlo.batch_norm_inference"(%183, %cst_20, %cst_21, %cst_21, %cst_20) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x2x2x1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>, tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %185 = xla_hlo.add %173, %184 : tensor<1x2x2x1024xf32> | |
| %186 = "xla_hlo.maximum"(%185, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x2x2x1024xf32>, tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %187 = "xla_hlo.convolution"(%186, %cst_23) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %188 = "xla_hlo.add"(%187, %cst_30) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %189 = "xla_hlo.batch_norm_inference"(%188, %cst_29, %cst_30, %cst_30, %cst_29) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %190 = "xla_hlo.convolution"(%186, %cst_24) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> | |
| %191 = "xla_hlo.add"(%190, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %192 = "xla_hlo.batch_norm_inference"(%191, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %193 = "xla_hlo.maximum"(%192, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %194 = "xla_hlo.convolution"(%193, %cst_28) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %195 = "xla_hlo.add"(%194, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %196 = "xla_hlo.batch_norm_inference"(%195, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %197 = "xla_hlo.maximum"(%196, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %198 = "xla_hlo.convolution"(%197, %cst_31) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %199 = "xla_hlo.add"(%198, %cst_30) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %200 = "xla_hlo.batch_norm_inference"(%199, %cst_29, %cst_30, %cst_30, %cst_29) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %201 = xla_hlo.add %189, %200 : tensor<1x1x1x2048xf32> | |
| %202 = "xla_hlo.maximum"(%201, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x2048xf32>, tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %203 = "xla_hlo.convolution"(%202, %cst_25) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| %204 = "xla_hlo.add"(%203, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %205 = "xla_hlo.batch_norm_inference"(%204, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %206 = "xla_hlo.maximum"(%205, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %207 = "xla_hlo.convolution"(%206, %cst_28) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %208 = "xla_hlo.add"(%207, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %209 = "xla_hlo.batch_norm_inference"(%208, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %210 = "xla_hlo.maximum"(%209, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %211 = "xla_hlo.convolution"(%210, %cst_31) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %212 = "xla_hlo.add"(%211, %cst_30) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %213 = "xla_hlo.batch_norm_inference"(%212, %cst_29, %cst_30, %cst_30, %cst_29) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %214 = xla_hlo.add %202, %213 : tensor<1x1x1x2048xf32> | |
| %215 = "xla_hlo.maximum"(%214, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x2048xf32>, tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %216 = "xla_hlo.convolution"(%215, %cst_25) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| %217 = "xla_hlo.add"(%216, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %218 = "xla_hlo.batch_norm_inference"(%217, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %219 = "xla_hlo.maximum"(%218, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %220 = "xla_hlo.convolution"(%219, %cst_28) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %221 = "xla_hlo.add"(%220, %cst_27) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %222 = "xla_hlo.batch_norm_inference"(%221, %cst_26, %cst_27, %cst_27, %cst_26) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>, tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %223 = "xla_hlo.maximum"(%222, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x512xf32>, tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %224 = "xla_hlo.convolution"(%223, %cst_31) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %225 = "xla_hlo.add"(%224, %cst_30) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %226 = "xla_hlo.batch_norm_inference"(%225, %cst_29, %cst_30, %cst_30, %cst_29) {epsilon = 1.001000e-05 : f32, feature_index = 3 : i64} : (tensor<1x1x1x2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>, tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %227 = xla_hlo.add %215, %226 : tensor<1x1x1x2048xf32> | |
| %228 = "xla_hlo.maximum"(%227, %1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x1x1x2048xf32>, tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| return %228 : tensor<1x1x1x2048xf32> | |
| } | |
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| func @predict() -> tensor<1x1x1x2048xf32> attributes {iree.module.export, iree.reflection = {abi = "sip", abiv = 1 : i32, f = "I1!R16!B12!d1d1d1d2048", fv = "1", sip = "I8!S5!k0_0R3!_0"}, tf._input_shapes = ["tfshape$dim { size: 1 } dim { size: 32 } dim { size: 32 } dim { size: 3 }", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true"], tf.signature.is_stateful} { | |
| %c16384 = constant 16384 : index | |
| %c20736 = constant 20736 : index | |
| %c4096 = constant 4096 : index | |
| %c8192 = constant 8192 : index | |
| %c2048 = constant 2048 : index | |
| %c1024 = constant 1024 : index | |
| %c512 = constant 512 : index | |
| %cst = constant dense<1.200000e+00> : tensor<1x32x32x3xf32> | |
| %cst_0 = constant dense<1.200000e+00> : tensor<7x7x3x64xf32> | |
| %cst_1 = constant dense<1.200000e+00> : tensor<1x1x64x64xf32> | |
| %cst_2 = constant dense<1.200000e+00> : tensor<1x1x256x64xf32> | |
| %cst_3 = constant dense<1.200000e+00> : tensor<3x3x64x64xf32> | |
| %cst_4 = constant dense<1.200000e+00> : tensor<1x1x64x256xf32> | |
| %cst_5 = constant dense<1.200000e+00> : tensor<1x1x256x512xf32> | |
| %cst_6 = constant dense<1.200000e+00> : tensor<1x1x256x128xf32> | |
| %cst_7 = constant dense<1.200000e+00> : tensor<1x1x512x128xf32> | |
| %cst_8 = constant dense<1.200000e+00> : tensor<3x3x128x128xf32> | |
| %cst_9 = constant dense<1.200000e+00> : tensor<1x1x128x512xf32> | |
| %cst_10 = constant dense<1.200000e+00> : tensor<1x1x512x1024xf32> | |
| %cst_11 = constant dense<1.200000e+00> : tensor<1x1x512x256xf32> | |
| %cst_12 = constant dense<1.200000e+00> : tensor<1x1x1024x256xf32> | |
| %cst_13 = constant dense<1.200000e+00> : tensor<3x3x256x256xf32> | |
| %cst_14 = constant dense<1.200000e+00> : tensor<1x1x256x1024xf32> | |
| %cst_15 = constant dense<1.200000e+00> : tensor<1x1x1024x2048xf32> | |
| %cst_16 = constant dense<1.200000e+00> : tensor<1x1x1024x512xf32> | |
| %cst_17 = constant dense<1.200000e+00> : tensor<1x1x2048x512xf32> | |
| %cst_18 = constant dense<1.200000e+00> : tensor<3x3x512x512xf32> | |
| %cst_19 = constant dense<1.200000e+00> : tensor<1x1x512x2048xf32> | |
| %0 = iree.do_not_optimize(%cst) : tensor<1x32x32x3xf32> | |
| %1 = flow.dispatch.region[%c16384 : index](%arg0 = %0 : tensor<1x32x32x3xf32>, %arg1 = %cst_0 : tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<3> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x32x32x3xf32>, tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> | |
| flow.return %105 : tensor<1x16x16x64xf32> | |
| } | |
| %2:4 = flow.dispatch.region[%c16384 : index](%arg0 = %1 : tensor<1x16x16x64xf32>) -> (tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x16x16x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x16x16x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x16x16x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x16x16x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x16x16x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x16x16x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x16x16x64xf32> | |
| %118 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %119 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %120 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| flow.return %117, %118, %119, %120 : tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32> | |
| } | |
| %3 = flow.dispatch.region[%c20736 : index](%arg0 = %2#0 : tensor<1x16x16x64xf32>) -> tensor<1x18x18x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %105 = "xla_hlo.pad"(%arg0, %cst_20) {edge_padding_high = dense<[0, 1, 1, 0]> : tensor<4xi64>, edge_padding_low = dense<[0, 1, 1, 0]> : tensor<4xi64>, interior_padding = dense<0> : tensor<4xi64>} : (tensor<1x16x16x64xf32>, tensor<f32>) -> tensor<1x18x18x64xf32> | |
| flow.return %105 : tensor<1x18x18x64xf32> | |
| } | |
| %4 = flow.dispatch.region[%c4096 : index](%arg0 = %3 : tensor<1x18x18x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0xFF800000> : tensor<f32> | |
| %105 = "xla_hlo.reduce_window"(%arg0, %cst_20) ( { | |
| ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>): // no predecessors | |
| %106 = xla_hlo.maximum %arg1, %arg2 : tensor<f32> | |
| "xla_hlo.return"(%106) : (tensor<f32>) -> () | |
| }) {window_dimensions = dense<[1, 3, 3, 1]> : tensor<4xi64>, window_strides = dense<[1, 2, 2, 1]> : tensor<4xi64>} : (tensor<1x18x18x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %5 = flow.dispatch.region[%c16384 : index](%arg0 = %4 : tensor<1x8x8x64xf32>, %arg1 = %cst_4 : tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| flow.return %105 : tensor<1x8x8x256xf32> | |
| } | |
| %6 = flow.dispatch.region[%c4096 : index](%arg0 = %4 : tensor<1x8x8x64xf32>, %arg1 = %cst_1 : tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %7:7 = flow.dispatch.region[%c4096 : index](%arg0 = %6 : tensor<1x8x8x64xf32>) -> (tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| %118 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %119 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %120 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %121 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %122 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %123 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %117, %118, %119, %120, %121, %122, %123 : tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32> | |
| } | |
| %8 = flow.dispatch.region[%c4096 : index](%arg0 = %7#0 : tensor<1x8x8x64xf32>, %arg1 = %cst_3 : tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %9 = flow.dispatch.region[%c4096 : index](%arg0 = %8 : tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| flow.return %117 : tensor<1x8x8x64xf32> | |
| } | |
| %10 = flow.dispatch.region[%c16384 : index](%arg0 = %9 : tensor<1x8x8x64xf32>, %arg1 = %cst_4 : tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| flow.return %105 : tensor<1x8x8x256xf32> | |
| } | |
| %11 = flow.dispatch.region[%c16384 : index](%arg0 = %5 : tensor<1x8x8x256xf32>, %arg1 = %10 : tensor<1x8x8x256xf32>, %arg2 = %2#1 : tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x256xf32> | |
| %116 = xla_hlo.add %arg1, %105 : tensor<1x8x8x256xf32> | |
| %117 = xla_hlo.subtract %116, %105 : tensor<1x8x8x256xf32> | |
| %118 = xla_hlo.multiply %117, %108 : tensor<1x8x8x256xf32> | |
| %119 = xla_hlo.divide %118, %113 : tensor<1x8x8x256xf32> | |
| %120 = xla_hlo.add %119, %105 : tensor<1x8x8x256xf32> | |
| %121 = xla_hlo.add %115, %120 : tensor<1x8x8x256xf32> | |
| %122 = xla_hlo.maximum %121, %arg2 : tensor<1x8x8x256xf32> | |
| flow.return %122 : tensor<1x8x8x256xf32> | |
| } | |
| %12 = flow.dispatch.region[%c4096 : index](%arg0 = %11 : tensor<1x8x8x256xf32>, %arg1 = %cst_2 : tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %13 = flow.dispatch.region[%c4096 : index](%arg0 = %12 : tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| flow.return %117 : tensor<1x8x8x64xf32> | |
| } | |
| %14 = flow.dispatch.region[%c4096 : index](%arg0 = %13 : tensor<1x8x8x64xf32>, %arg1 = %cst_3 : tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %15 = flow.dispatch.region[%c4096 : index](%arg0 = %14 : tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| flow.return %117 : tensor<1x8x8x64xf32> | |
| } | |
| %16 = flow.dispatch.region[%c16384 : index](%arg0 = %15 : tensor<1x8x8x64xf32>, %arg1 = %cst_4 : tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| flow.return %105 : tensor<1x8x8x256xf32> | |
| } | |
| %17 = flow.dispatch.region[%c16384 : index](%arg0 = %16 : tensor<1x8x8x256xf32>, %arg1 = %11 : tensor<1x8x8x256xf32>, %arg2 = %2#2 : tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x256xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x8x8x256xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x8x8x256xf32> | |
| flow.return %117 : tensor<1x8x8x256xf32> | |
| } | |
| %18 = flow.dispatch.region[%c4096 : index](%arg0 = %17 : tensor<1x8x8x256xf32>, %arg1 = %cst_2 : tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %19 = flow.dispatch.region[%c4096 : index](%arg0 = %18 : tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| flow.return %117 : tensor<1x8x8x64xf32> | |
| } | |
| %20 = flow.dispatch.region[%c4096 : index](%arg0 = %19 : tensor<1x8x8x64xf32>, %arg1 = %cst_3 : tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| flow.return %105 : tensor<1x8x8x64xf32> | |
| } | |
| %21 = flow.dispatch.region[%c4096 : index](%arg0 = %20 : tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x64xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x64xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x64xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<64xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<64xf32>) -> tensor<64xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x64xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x64xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x8x8x64xf32> | |
| flow.return %117 : tensor<1x8x8x64xf32> | |
| } | |
| %22 = flow.dispatch.region[%c16384 : index](%arg0 = %21 : tensor<1x8x8x64xf32>, %arg1 = %cst_4 : tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| flow.return %105 : tensor<1x8x8x256xf32> | |
| } | |
| %23 = flow.dispatch.region[%c16384 : index](%arg0 = %22 : tensor<1x8x8x256xf32>, %arg1 = %17 : tensor<1x8x8x256xf32>, %arg2 = %2#3 : tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x8x8x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x8x8x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x8x8x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x8x8x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x8x8x256xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x8x8x256xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x8x8x256xf32> | |
| flow.return %117 : tensor<1x8x8x256xf32> | |
| } | |
| %24 = flow.dispatch.region[%c8192 : index](%arg0 = %23 : tensor<1x8x8x256xf32>, %arg1 = %cst_5 : tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> | |
| flow.return %105 : tensor<1x4x4x512xf32> | |
| } | |
| %25 = flow.dispatch.region[%c2048 : index](%arg0 = %23 : tensor<1x8x8x256xf32>, %arg1 = %cst_6 : tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %26:3 = flow.dispatch.region[%c2048 : index](%arg0 = %25 : tensor<1x4x4x128xf32>) -> (tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| %118 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %119 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %117, %118, %119 : tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32> | |
| } | |
| %27 = flow.dispatch.region[%c2048 : index](%arg0 = %26#0 : tensor<1x4x4x128xf32>, %arg1 = %cst_8 : tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %28 = flow.dispatch.region[%c2048 : index](%arg0 = %27 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %29 = flow.dispatch.region[%c8192 : index](%arg0 = %28 : tensor<1x4x4x128xf32>, %arg1 = %cst_9 : tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| flow.return %105 : tensor<1x4x4x512xf32> | |
| } | |
| %30:4 = flow.dispatch.region[%c8192 : index](%arg0 = %24 : tensor<1x4x4x512xf32>, %arg1 = %29 : tensor<1x4x4x512xf32>) -> (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x512xf32> | |
| %116 = xla_hlo.add %arg1, %105 : tensor<1x4x4x512xf32> | |
| %117 = xla_hlo.subtract %116, %105 : tensor<1x4x4x512xf32> | |
| %118 = xla_hlo.multiply %117, %108 : tensor<1x4x4x512xf32> | |
| %119 = xla_hlo.divide %118, %113 : tensor<1x4x4x512xf32> | |
| %120 = xla_hlo.add %119, %105 : tensor<1x4x4x512xf32> | |
| %121 = xla_hlo.add %115, %120 : tensor<1x4x4x512xf32> | |
| %122 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %123 = xla_hlo.maximum %121, %122 : tensor<1x4x4x512xf32> | |
| %124 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %125 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %126 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| flow.return %123, %124, %125, %126 : tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32> | |
| } | |
| %31 = flow.dispatch.region[%c2048 : index](%arg0 = %30#0 : tensor<1x4x4x512xf32>, %arg1 = %cst_7 : tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %32 = flow.dispatch.region[%c2048 : index](%arg0 = %31 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %33 = flow.dispatch.region[%c2048 : index](%arg0 = %32 : tensor<1x4x4x128xf32>, %arg1 = %cst_8 : tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %34 = flow.dispatch.region[%c2048 : index](%arg0 = %33 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %35 = flow.dispatch.region[%c8192 : index](%arg0 = %34 : tensor<1x4x4x128xf32>, %arg1 = %cst_9 : tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| flow.return %105 : tensor<1x4x4x512xf32> | |
| } | |
| %36 = flow.dispatch.region[%c8192 : index](%arg0 = %35 : tensor<1x4x4x512xf32>, %arg1 = %30#0 : tensor<1x4x4x512xf32>, %arg2 = %30#1 : tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x512xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x4x4x512xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x4x4x512xf32> | |
| flow.return %117 : tensor<1x4x4x512xf32> | |
| } | |
| %37 = flow.dispatch.region[%c2048 : index](%arg0 = %36 : tensor<1x4x4x512xf32>, %arg1 = %cst_7 : tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %38 = flow.dispatch.region[%c2048 : index](%arg0 = %37 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %39 = flow.dispatch.region[%c2048 : index](%arg0 = %38 : tensor<1x4x4x128xf32>, %arg1 = %cst_8 : tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %40 = flow.dispatch.region[%c2048 : index](%arg0 = %39 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %41 = flow.dispatch.region[%c8192 : index](%arg0 = %40 : tensor<1x4x4x128xf32>, %arg1 = %cst_9 : tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| flow.return %105 : tensor<1x4x4x512xf32> | |
| } | |
| %42 = flow.dispatch.region[%c8192 : index](%arg0 = %41 : tensor<1x4x4x512xf32>, %arg1 = %36 : tensor<1x4x4x512xf32>, %arg2 = %30#2 : tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x512xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x4x4x512xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x4x4x512xf32> | |
| flow.return %117 : tensor<1x4x4x512xf32> | |
| } | |
| %43 = flow.dispatch.region[%c2048 : index](%arg0 = %42 : tensor<1x4x4x512xf32>, %arg1 = %cst_7 : tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %44 = flow.dispatch.region[%c2048 : index](%arg0 = %43 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %45 = flow.dispatch.region[%c2048 : index](%arg0 = %44 : tensor<1x4x4x128xf32>, %arg1 = %cst_8 : tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| flow.return %105 : tensor<1x4x4x128xf32> | |
| } | |
| %46 = flow.dispatch.region[%c2048 : index](%arg0 = %45 : tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x128xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x128xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x128xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<128xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<128xf32>) -> tensor<128xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x128xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x128xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x4x4x128xf32> | |
| flow.return %117 : tensor<1x4x4x128xf32> | |
| } | |
| %47 = flow.dispatch.region[%c8192 : index](%arg0 = %46 : tensor<1x4x4x128xf32>, %arg1 = %cst_9 : tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| flow.return %105 : tensor<1x4x4x512xf32> | |
| } | |
| %48 = flow.dispatch.region[%c8192 : index](%arg0 = %47 : tensor<1x4x4x512xf32>, %arg1 = %42 : tensor<1x4x4x512xf32>, %arg2 = %30#3 : tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x4x4x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x4x4x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x4x4x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x4x4x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x4x4x512xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x4x4x512xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x4x4x512xf32> | |
| flow.return %117 : tensor<1x4x4x512xf32> | |
| } | |
| %49 = flow.dispatch.region[%c4096 : index](%arg0 = %48 : tensor<1x4x4x512xf32>, %arg1 = %cst_10 : tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %50 = flow.dispatch.region[%c1024 : index](%arg0 = %48 : tensor<1x4x4x512xf32>, %arg1 = %cst_11 : tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %51 = flow.dispatch.region[%c1024 : index](%arg0 = %50 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %52 = flow.dispatch.region[%c1024 : index](%arg0 = %51 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %53 = flow.dispatch.region[%c1024 : index](%arg0 = %52 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %54 = flow.dispatch.region[%c4096 : index](%arg0 = %53 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %55 = flow.dispatch.region[%c4096 : index](%arg0 = %49 : tensor<1x2x2x1024xf32>, %arg1 = %54 : tensor<1x2x2x1024xf32>, %arg2 = %7#1 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %105 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.subtract %116, %105 : tensor<1x2x2x1024xf32> | |
| %118 = xla_hlo.multiply %117, %108 : tensor<1x2x2x1024xf32> | |
| %119 = xla_hlo.divide %118, %113 : tensor<1x2x2x1024xf32> | |
| %120 = xla_hlo.add %119, %105 : tensor<1x2x2x1024xf32> | |
| %121 = xla_hlo.add %115, %120 : tensor<1x2x2x1024xf32> | |
| %122 = xla_hlo.maximum %121, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %122 : tensor<1x2x2x1024xf32> | |
| } | |
| %56 = flow.dispatch.region[%c1024 : index](%arg0 = %55 : tensor<1x2x2x1024xf32>, %arg1 = %cst_12 : tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %57 = flow.dispatch.region[%c1024 : index](%arg0 = %56 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %58 = flow.dispatch.region[%c1024 : index](%arg0 = %57 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %59 = flow.dispatch.region[%c1024 : index](%arg0 = %58 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %60 = flow.dispatch.region[%c4096 : index](%arg0 = %59 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %61 = flow.dispatch.region[%c4096 : index](%arg0 = %60 : tensor<1x2x2x1024xf32>, %arg1 = %55 : tensor<1x2x2x1024xf32>, %arg2 = %7#2 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %117 : tensor<1x2x2x1024xf32> | |
| } | |
| %62 = flow.dispatch.region[%c1024 : index](%arg0 = %61 : tensor<1x2x2x1024xf32>, %arg1 = %cst_12 : tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %63 = flow.dispatch.region[%c1024 : index](%arg0 = %62 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %64 = flow.dispatch.region[%c1024 : index](%arg0 = %63 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %65 = flow.dispatch.region[%c1024 : index](%arg0 = %64 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %66 = flow.dispatch.region[%c4096 : index](%arg0 = %65 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %67 = flow.dispatch.region[%c4096 : index](%arg0 = %66 : tensor<1x2x2x1024xf32>, %arg1 = %61 : tensor<1x2x2x1024xf32>, %arg2 = %7#3 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %117 : tensor<1x2x2x1024xf32> | |
| } | |
| %68 = flow.dispatch.region[%c1024 : index](%arg0 = %67 : tensor<1x2x2x1024xf32>, %arg1 = %cst_12 : tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %69 = flow.dispatch.region[%c1024 : index](%arg0 = %68 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %70 = flow.dispatch.region[%c1024 : index](%arg0 = %69 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %71 = flow.dispatch.region[%c1024 : index](%arg0 = %70 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %72 = flow.dispatch.region[%c4096 : index](%arg0 = %71 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %73 = flow.dispatch.region[%c4096 : index](%arg0 = %72 : tensor<1x2x2x1024xf32>, %arg1 = %67 : tensor<1x2x2x1024xf32>, %arg2 = %7#4 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %117 : tensor<1x2x2x1024xf32> | |
| } | |
| %74 = flow.dispatch.region[%c1024 : index](%arg0 = %73 : tensor<1x2x2x1024xf32>, %arg1 = %cst_12 : tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %75 = flow.dispatch.region[%c1024 : index](%arg0 = %74 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %76 = flow.dispatch.region[%c1024 : index](%arg0 = %75 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %77 = flow.dispatch.region[%c1024 : index](%arg0 = %76 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %78 = flow.dispatch.region[%c4096 : index](%arg0 = %77 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %79 = flow.dispatch.region[%c4096 : index](%arg0 = %78 : tensor<1x2x2x1024xf32>, %arg1 = %73 : tensor<1x2x2x1024xf32>, %arg2 = %7#5 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %117 : tensor<1x2x2x1024xf32> | |
| } | |
| %80 = flow.dispatch.region[%c1024 : index](%arg0 = %79 : tensor<1x2x2x1024xf32>, %arg1 = %cst_12 : tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %81 = flow.dispatch.region[%c1024 : index](%arg0 = %80 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %82 = flow.dispatch.region[%c1024 : index](%arg0 = %81 : tensor<1x2x2x256xf32>, %arg1 = %cst_13 : tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| flow.return %105 : tensor<1x2x2x256xf32> | |
| } | |
| %83 = flow.dispatch.region[%c1024 : index](%arg0 = %82 : tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x256xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x256xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x256xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<256xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<256xf32>) -> tensor<256xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x256xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x256xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x2x2x256xf32> | |
| flow.return %117 : tensor<1x2x2x256xf32> | |
| } | |
| %84 = flow.dispatch.region[%c4096 : index](%arg0 = %83 : tensor<1x2x2x256xf32>, %arg1 = %cst_14 : tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| flow.return %105 : tensor<1x2x2x1024xf32> | |
| } | |
| %85 = flow.dispatch.region[%c4096 : index](%arg0 = %84 : tensor<1x2x2x1024xf32>, %arg1 = %79 : tensor<1x2x2x1024xf32>, %arg2 = %7#6 : tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x2x2x1024xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x2x2x1024xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x2x2x1024xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<1024xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x2x2x1024xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x2x2x1024xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x2x2x1024xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x2x2x1024xf32> | |
| flow.return %117 : tensor<1x2x2x1024xf32> | |
| } | |
| %86 = flow.dispatch.region[%c2048 : index](%arg0 = %85 : tensor<1x2x2x1024xf32>, %arg1 = %cst_15 : tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %105 : tensor<1x1x1x2048xf32> | |
| } | |
| %87 = flow.dispatch.region[%c512 : index](%arg0 = %85 : tensor<1x2x2x1024xf32>, %arg1 = %cst_16 : tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %88 = flow.dispatch.region[%c512 : index](%arg0 = %87 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %89 = flow.dispatch.region[%c512 : index](%arg0 = %88 : tensor<1x1x1x512xf32>, %arg1 = %cst_18 : tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %90 = flow.dispatch.region[%c512 : index](%arg0 = %89 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %91 = flow.dispatch.region[%c2048 : index](%arg0 = %90 : tensor<1x1x1x512xf32>, %arg1 = %cst_19 : tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %105 : tensor<1x1x1x2048xf32> | |
| } | |
| %92 = flow.dispatch.region[%c2048 : index](%arg0 = %86 : tensor<1x1x1x2048xf32>, %arg1 = %91 : tensor<1x1x1x2048xf32>, %arg2 = %26#1 : tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x2048xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x2048xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x2048xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<2048xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x2048xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x2048xf32> | |
| %116 = xla_hlo.add %arg1, %105 : tensor<1x1x1x2048xf32> | |
| %117 = xla_hlo.subtract %116, %105 : tensor<1x1x1x2048xf32> | |
| %118 = xla_hlo.multiply %117, %108 : tensor<1x1x1x2048xf32> | |
| %119 = xla_hlo.divide %118, %113 : tensor<1x1x1x2048xf32> | |
| %120 = xla_hlo.add %119, %105 : tensor<1x1x1x2048xf32> | |
| %121 = xla_hlo.add %115, %120 : tensor<1x1x1x2048xf32> | |
| %122 = xla_hlo.maximum %121, %arg2 : tensor<1x1x1x2048xf32> | |
| flow.return %122 : tensor<1x1x1x2048xf32> | |
| } | |
| %93 = flow.dispatch.region[%c512 : index](%arg0 = %92 : tensor<1x1x1x2048xf32>, %arg1 = %cst_17 : tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %94 = flow.dispatch.region[%c512 : index](%arg0 = %93 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %95 = flow.dispatch.region[%c512 : index](%arg0 = %94 : tensor<1x1x1x512xf32>, %arg1 = %cst_18 : tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %96 = flow.dispatch.region[%c512 : index](%arg0 = %95 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %97 = flow.dispatch.region[%c2048 : index](%arg0 = %96 : tensor<1x1x1x512xf32>, %arg1 = %cst_19 : tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %105 : tensor<1x1x1x2048xf32> | |
| } | |
| %98 = flow.dispatch.region[%c2048 : index](%arg0 = %97 : tensor<1x1x1x2048xf32>, %arg1 = %92 : tensor<1x1x1x2048xf32>, %arg2 = %26#2 : tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst_20 = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_21 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_22 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x2048xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x2048xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x2048xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %111 = xla_hlo.add %110, %cst_20 : tensor<2048xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x2048xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x2048xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x1x1x2048xf32> | |
| %117 = xla_hlo.maximum %116, %arg2 : tensor<1x1x1x2048xf32> | |
| flow.return %117 : tensor<1x1x1x2048xf32> | |
| } | |
| %99 = flow.dispatch.region[%c512 : index](%arg0 = %98 : tensor<1x1x1x2048xf32>, %arg1 = %cst_17 : tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %100 = flow.dispatch.region[%c512 : index](%arg0 = %99 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %101 = flow.dispatch.region[%c512 : index](%arg0 = %100 : tensor<1x1x1x512xf32>, %arg1 = %cst_18 : tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| flow.return %105 : tensor<1x1x1x512xf32> | |
| } | |
| %102 = flow.dispatch.region[%c512 : index](%arg0 = %101 : tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x512xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x512xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x512xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<512xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<512xf32>) -> tensor<512xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x512xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x512xf32> | |
| %116 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %117 = xla_hlo.maximum %115, %116 : tensor<1x1x1x512xf32> | |
| flow.return %117 : tensor<1x1x1x512xf32> | |
| } | |
| %103 = flow.dispatch.region[%c2048 : index](%arg0 = %102 : tensor<1x1x1x512xf32>, %arg1 = %cst_19 : tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %105 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %105 : tensor<1x1x1x2048xf32> | |
| } | |
| %104 = flow.dispatch.region[%c2048 : index](%arg0 = %103 : tensor<1x1x1x2048xf32>, %arg1 = %98 : tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst_20 = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_21 = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_22 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_23 = constant dense<1.001000e-05> : tensor<f32> | |
| %105 = "xla_hlo.broadcast_in_dim"(%cst_22) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %106 = xla_hlo.add %arg0, %105 : tensor<1x1x1x2048xf32> | |
| %107 = xla_hlo.subtract %106, %105 : tensor<1x1x1x2048xf32> | |
| %108 = "xla_hlo.broadcast_in_dim"(%cst_21) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %109 = xla_hlo.multiply %107, %108 : tensor<1x1x1x2048xf32> | |
| %110 = "xla_hlo.broadcast_in_dim"(%cst_23) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %111 = xla_hlo.add %110, %cst_21 : tensor<2048xf32> | |
| %112 = "xla_hlo.sqrt"(%111) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %113 = "xla_hlo.broadcast_in_dim"(%112) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %114 = xla_hlo.divide %109, %113 : tensor<1x1x1x2048xf32> | |
| %115 = xla_hlo.add %114, %105 : tensor<1x1x1x2048xf32> | |
| %116 = xla_hlo.add %arg1, %115 : tensor<1x1x1x2048xf32> | |
| %117 = "xla_hlo.broadcast_in_dim"(%cst_20) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %118 = xla_hlo.maximum %116, %117 : tensor<1x1x1x2048xf32> | |
| flow.return %118 : tensor<1x1x1x2048xf32> | |
| } | |
| return %104 : tensor<1x1x1x2048xf32> | |
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 370 : i32}} { | |
| flow.executable @predict_ex_dispatch_0 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_0 | |
| module { | |
| func @predict_ex_dispatch_0(%arg0: tensor<1x32x32x3xf32>, %arg1: tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<3> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x32x32x3xf32>, tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> | |
| return %0 : tensor<1x16x16x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_1 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_1 | |
| module { | |
| func @predict_ex_dispatch_1(%arg0: tensor<1x16x16x64xf32>) -> (tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x16x16x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x16x16x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x16x16x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x16x16x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x16x16x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x16x16x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x16x16x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x16x16x64xf32> | |
| %13 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %14 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| %15 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x256xf32> | |
| return %12, %13, %14, %15 : tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_2 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_2 | |
| module { | |
| func @predict_ex_dispatch_2(%arg0: tensor<1x16x16x64xf32>) -> tensor<1x18x18x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %0 = "xla_hlo.pad"(%arg0, %cst) {edge_padding_high = dense<[0, 1, 1, 0]> : tensor<4xi64>, edge_padding_low = dense<[0, 1, 1, 0]> : tensor<4xi64>, interior_padding = dense<0> : tensor<4xi64>} : (tensor<1x16x16x64xf32>, tensor<f32>) -> tensor<1x18x18x64xf32> | |
| return %0 : tensor<1x18x18x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_3 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_3 | |
| module { | |
| func @predict_ex_dispatch_3(%arg0: tensor<1x18x18x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0xFF800000> : tensor<f32> | |
| %0 = "xla_hlo.reduce_window"(%arg0, %cst) ( { | |
| ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>): // no predecessors | |
| %1 = xla_hlo.maximum %arg1, %arg2 : tensor<f32> | |
| "xla_hlo.return"(%1) : (tensor<f32>) -> () | |
| }) {window_dimensions = dense<[1, 3, 3, 1]> : tensor<4xi64>, window_strides = dense<[1, 2, 2, 1]> : tensor<4xi64>} : (tensor<1x18x18x64xf32>, tensor<f32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_4 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_4 | |
| module { | |
| func @predict_ex_dispatch_4(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| return %0 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_5 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_5 | |
| module { | |
| func @predict_ex_dispatch_5(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_6 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_6 | |
| module { | |
| func @predict_ex_dispatch_6(%arg0: tensor<1x8x8x64xf32>) -> (tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| %13 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %14 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %15 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %16 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %17 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| %18 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x1024xf32> | |
| return %12, %13, %14, %15, %16, %17, %18 : tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_7 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_7 | |
| module { | |
| func @predict_ex_dispatch_7(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_8 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_8 | |
| module { | |
| func @predict_ex_dispatch_8(%arg0: tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| return %12 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_9 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_9 | |
| module { | |
| func @predict_ex_dispatch_9(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| return %0 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_10 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_10 | |
| module { | |
| func @predict_ex_dispatch_10(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x8x8x256xf32>, %arg2: tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x256xf32> | |
| %11 = xla_hlo.add %arg1, %0 : tensor<1x8x8x256xf32> | |
| %12 = xla_hlo.subtract %11, %0 : tensor<1x8x8x256xf32> | |
| %13 = xla_hlo.multiply %12, %3 : tensor<1x8x8x256xf32> | |
| %14 = xla_hlo.divide %13, %8 : tensor<1x8x8x256xf32> | |
| %15 = xla_hlo.add %14, %0 : tensor<1x8x8x256xf32> | |
| %16 = xla_hlo.add %10, %15 : tensor<1x8x8x256xf32> | |
| %17 = xla_hlo.maximum %16, %arg2 : tensor<1x8x8x256xf32> | |
| return %17 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_11 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_11 | |
| module { | |
| func @predict_ex_dispatch_11(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_12 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_12 | |
| module { | |
| func @predict_ex_dispatch_12(%arg0: tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| return %12 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_13 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_13 | |
| module { | |
| func @predict_ex_dispatch_13(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_14 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_14 | |
| module { | |
| func @predict_ex_dispatch_14(%arg0: tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| return %12 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_15 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_15 | |
| module { | |
| func @predict_ex_dispatch_15(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| return %0 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_16 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_16 | |
| module { | |
| func @predict_ex_dispatch_16(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x8x8x256xf32>, %arg2: tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x256xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x8x8x256xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x8x8x256xf32> | |
| return %12 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_17 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_17 | |
| module { | |
| func @predict_ex_dispatch_17(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_18 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_18 | |
| module { | |
| func @predict_ex_dispatch_18(%arg0: tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| return %12 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_19 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_19 | |
| module { | |
| func @predict_ex_dispatch_19(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| return %0 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_20 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_20 | |
| module { | |
| func @predict_ex_dispatch_20(%arg0: tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<64xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<64xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x64xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x64xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x64xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<64xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<64xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<64xf32>) -> tensor<64xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<64xf32>) -> tensor<1x8x8x64xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x64xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x64xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x8x8x64xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x8x8x64xf32> | |
| return %12 : tensor<1x8x8x64xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_21 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_21 | |
| module { | |
| func @predict_ex_dispatch_21(%arg0: tensor<1x8x8x64xf32>, %arg1: tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| return %0 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_22 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_22 | |
| module { | |
| func @predict_ex_dispatch_22(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x8x8x256xf32>, %arg2: tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x8x8x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x8x8x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x8x8x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x8x8x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x8x8x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x8x8x256xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x8x8x256xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x8x8x256xf32> | |
| return %12 : tensor<1x8x8x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_23 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_23 | |
| module { | |
| func @predict_ex_dispatch_23(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> | |
| return %0 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_24 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_24 | |
| module { | |
| func @predict_ex_dispatch_24(%arg0: tensor<1x8x8x256xf32>, %arg1: tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x8x8x256xf32>, tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_25 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_25 | |
| module { | |
| func @predict_ex_dispatch_25(%arg0: tensor<1x4x4x128xf32>) -> (tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| %13 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %14 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| return %12, %13, %14 : tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_26 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_26 | |
| module { | |
| func @predict_ex_dispatch_26(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_27 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_27 | |
| module { | |
| func @predict_ex_dispatch_27(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_28 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_28 | |
| module { | |
| func @predict_ex_dispatch_28(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| return %0 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_29 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_29 | |
| module { | |
| func @predict_ex_dispatch_29(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x4x4x512xf32>) -> (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x512xf32> | |
| %11 = xla_hlo.add %arg1, %0 : tensor<1x4x4x512xf32> | |
| %12 = xla_hlo.subtract %11, %0 : tensor<1x4x4x512xf32> | |
| %13 = xla_hlo.multiply %12, %3 : tensor<1x4x4x512xf32> | |
| %14 = xla_hlo.divide %13, %8 : tensor<1x4x4x512xf32> | |
| %15 = xla_hlo.add %14, %0 : tensor<1x4x4x512xf32> | |
| %16 = xla_hlo.add %10, %15 : tensor<1x4x4x512xf32> | |
| %17 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %18 = xla_hlo.maximum %16, %17 : tensor<1x4x4x512xf32> | |
| %19 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %20 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| %21 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x512xf32> | |
| return %18, %19, %20, %21 : tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_30 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_30 | |
| module { | |
| func @predict_ex_dispatch_30(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_31 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_31 | |
| module { | |
| func @predict_ex_dispatch_31(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_32 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_32 | |
| module { | |
| func @predict_ex_dispatch_32(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_33 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_33 | |
| module { | |
| func @predict_ex_dispatch_33(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_34 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_34 | |
| module { | |
| func @predict_ex_dispatch_34(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| return %0 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_35 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_35 | |
| module { | |
| func @predict_ex_dispatch_35(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x4x4x512xf32>, %arg2: tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x512xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x4x4x512xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x4x4x512xf32> | |
| return %12 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_36 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_36 | |
| module { | |
| func @predict_ex_dispatch_36(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_37 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_37 | |
| module { | |
| func @predict_ex_dispatch_37(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_38 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_38 | |
| module { | |
| func @predict_ex_dispatch_38(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_39 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_39 | |
| module { | |
| func @predict_ex_dispatch_39(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_40 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_40 | |
| module { | |
| func @predict_ex_dispatch_40(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| return %0 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_41 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_41 | |
| module { | |
| func @predict_ex_dispatch_41(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x4x4x512xf32>, %arg2: tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x512xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x4x4x512xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x4x4x512xf32> | |
| return %12 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_42 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_42 | |
| module { | |
| func @predict_ex_dispatch_42(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_43 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_43 | |
| module { | |
| func @predict_ex_dispatch_43(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_44 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_44 | |
| module { | |
| func @predict_ex_dispatch_44(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| return %0 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_45 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_45 | |
| module { | |
| func @predict_ex_dispatch_45(%arg0: tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<128xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<128xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x128xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x128xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x128xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<128xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<128xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<128xf32>) -> tensor<128xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x4x4x128xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x128xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x128xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x4x4x128xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x4x4x128xf32> | |
| return %12 : tensor<1x4x4x128xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_46 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_46 | |
| module { | |
| func @predict_ex_dispatch_46(%arg0: tensor<1x4x4x128xf32>, %arg1: tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| return %0 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_47 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_47 | |
| module { | |
| func @predict_ex_dispatch_47(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x4x4x512xf32>, %arg2: tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x4x4x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x4x4x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x4x4x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x4x4x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x4x4x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x4x4x512xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x4x4x512xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x4x4x512xf32> | |
| return %12 : tensor<1x4x4x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_48 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_48 | |
| module { | |
| func @predict_ex_dispatch_48(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_49 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_49 | |
| module { | |
| func @predict_ex_dispatch_49(%arg0: tensor<1x4x4x512xf32>, %arg1: tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x4x4x512xf32>, tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_50 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_50 | |
| module { | |
| func @predict_ex_dispatch_50(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_51 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_51 | |
| module { | |
| func @predict_ex_dispatch_51(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_52 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_52 | |
| module { | |
| func @predict_ex_dispatch_52(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_53 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_53 | |
| module { | |
| func @predict_ex_dispatch_53(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_54 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_54 | |
| module { | |
| func @predict_ex_dispatch_54(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %0 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.subtract %11, %0 : tensor<1x2x2x1024xf32> | |
| %13 = xla_hlo.multiply %12, %3 : tensor<1x2x2x1024xf32> | |
| %14 = xla_hlo.divide %13, %8 : tensor<1x2x2x1024xf32> | |
| %15 = xla_hlo.add %14, %0 : tensor<1x2x2x1024xf32> | |
| %16 = xla_hlo.add %10, %15 : tensor<1x2x2x1024xf32> | |
| %17 = xla_hlo.maximum %16, %arg2 : tensor<1x2x2x1024xf32> | |
| return %17 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_55 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_55 | |
| module { | |
| func @predict_ex_dispatch_55(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_56 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_56 | |
| module { | |
| func @predict_ex_dispatch_56(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_57 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_57 | |
| module { | |
| func @predict_ex_dispatch_57(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_58 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_58 | |
| module { | |
| func @predict_ex_dispatch_58(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_59 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_59 | |
| module { | |
| func @predict_ex_dispatch_59(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_60 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_60 | |
| module { | |
| func @predict_ex_dispatch_60(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x2x2x1024xf32> | |
| return %12 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_61 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_61 | |
| module { | |
| func @predict_ex_dispatch_61(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_62 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_62 | |
| module { | |
| func @predict_ex_dispatch_62(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_63 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_63 | |
| module { | |
| func @predict_ex_dispatch_63(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_64 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_64 | |
| module { | |
| func @predict_ex_dispatch_64(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_65 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_65 | |
| module { | |
| func @predict_ex_dispatch_65(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_66 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_66 | |
| module { | |
| func @predict_ex_dispatch_66(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x2x2x1024xf32> | |
| return %12 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_67 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_67 | |
| module { | |
| func @predict_ex_dispatch_67(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_68 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_68 | |
| module { | |
| func @predict_ex_dispatch_68(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_69 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_69 | |
| module { | |
| func @predict_ex_dispatch_69(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_70 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_70 | |
| module { | |
| func @predict_ex_dispatch_70(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_71 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_71 | |
| module { | |
| func @predict_ex_dispatch_71(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_72 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_72 | |
| module { | |
| func @predict_ex_dispatch_72(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x2x2x1024xf32> | |
| return %12 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_73 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_73 | |
| module { | |
| func @predict_ex_dispatch_73(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_74 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_74 | |
| module { | |
| func @predict_ex_dispatch_74(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_75 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_75 | |
| module { | |
| func @predict_ex_dispatch_75(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_76 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_76 | |
| module { | |
| func @predict_ex_dispatch_76(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_77 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_77 | |
| module { | |
| func @predict_ex_dispatch_77(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_78 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_78 | |
| module { | |
| func @predict_ex_dispatch_78(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x2x2x1024xf32> | |
| return %12 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_79 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_79 | |
| module { | |
| func @predict_ex_dispatch_79(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_80 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_80 | |
| module { | |
| func @predict_ex_dispatch_80(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_81 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_81 | |
| module { | |
| func @predict_ex_dispatch_81(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| return %0 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_82 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_82 | |
| module { | |
| func @predict_ex_dispatch_82(%arg0: tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<256xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<256xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x256xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x256xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x256xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<256xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<256xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<256xf32>) -> tensor<256xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<256xf32>) -> tensor<1x2x2x256xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x256xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x256xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x2x2x256xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x2x2x256xf32> | |
| return %12 : tensor<1x2x2x256xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_83 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_83 | |
| module { | |
| func @predict_ex_dispatch_83(%arg0: tensor<1x2x2x256xf32>, %arg1: tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| return %0 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_84 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_84 | |
| module { | |
| func @predict_ex_dispatch_84(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x2x2x1024xf32>, %arg2: tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<1024xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<1024xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x2x2x1024xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x2x2x1024xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x2x2x1024xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1024xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<1024xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<1024xf32>) -> tensor<1024xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x2x2x1024xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x2x2x1024xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x2x2x1024xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x2x2x1024xf32> | |
| return %12 : tensor<1x2x2x1024xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_85 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_85 | |
| module { | |
| func @predict_ex_dispatch_85(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| return %0 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_86 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_86 | |
| module { | |
| func @predict_ex_dispatch_86(%arg0: tensor<1x2x2x1024xf32>, %arg1: tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<2> : tensor<2xi64>} : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_87 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_87 | |
| module { | |
| func @predict_ex_dispatch_87(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_88 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_88 | |
| module { | |
| func @predict_ex_dispatch_88(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_89 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_89 | |
| module { | |
| func @predict_ex_dispatch_89(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_90 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_90 | |
| module { | |
| func @predict_ex_dispatch_90(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| return %0 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_91 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_91 | |
| module { | |
| func @predict_ex_dispatch_91(%arg0: tensor<1x1x1x2048xf32>, %arg1: tensor<1x1x1x2048xf32>, %arg2: tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x2048xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x2048xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x2048xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<2048xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x2048xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x2048xf32> | |
| %11 = xla_hlo.add %arg1, %0 : tensor<1x1x1x2048xf32> | |
| %12 = xla_hlo.subtract %11, %0 : tensor<1x1x1x2048xf32> | |
| %13 = xla_hlo.multiply %12, %3 : tensor<1x1x1x2048xf32> | |
| %14 = xla_hlo.divide %13, %8 : tensor<1x1x1x2048xf32> | |
| %15 = xla_hlo.add %14, %0 : tensor<1x1x1x2048xf32> | |
| %16 = xla_hlo.add %10, %15 : tensor<1x1x1x2048xf32> | |
| %17 = xla_hlo.maximum %16, %arg2 : tensor<1x1x1x2048xf32> | |
| return %17 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_92 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_92 | |
| module { | |
| func @predict_ex_dispatch_92(%arg0: tensor<1x1x1x2048xf32>, %arg1: tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_93 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_93 | |
| module { | |
| func @predict_ex_dispatch_93(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_94 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_94 | |
| module { | |
| func @predict_ex_dispatch_94(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_95 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_95 | |
| module { | |
| func @predict_ex_dispatch_95(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_96 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_96 | |
| module { | |
| func @predict_ex_dispatch_96(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| return %0 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_97 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_97 | |
| module { | |
| func @predict_ex_dispatch_97(%arg0: tensor<1x1x1x2048xf32>, %arg1: tensor<1x1x1x2048xf32>, %arg2: tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_0 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_1 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x2048xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x2048xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x2048xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %6 = xla_hlo.add %5, %cst : tensor<2048xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x2048xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x2048xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x1x1x2048xf32> | |
| %12 = xla_hlo.maximum %11, %arg2 : tensor<1x1x1x2048xf32> | |
| return %12 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_98 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_98 | |
| module { | |
| func @predict_ex_dispatch_98(%arg0: tensor<1x1x1x2048xf32>, %arg1: tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_99 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_99 | |
| module { | |
| func @predict_ex_dispatch_99(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_100 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_100 | |
| module { | |
| func @predict_ex_dispatch_100(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<1> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| return %0 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_101 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_101 | |
| module { | |
| func @predict_ex_dispatch_101(%arg0: tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<512xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<512xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x512xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x512xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x512xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<512xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<512xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<512xf32>) -> tensor<512xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<512xf32>) -> tensor<1x1x1x512xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x512xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x512xf32> | |
| %11 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x512xf32> | |
| %12 = xla_hlo.maximum %10, %11 : tensor<1x1x1x512xf32> | |
| return %12 : tensor<1x1x1x512xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_102 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_102 | |
| module { | |
| func @predict_ex_dispatch_102(%arg0: tensor<1x1x1x512xf32>, %arg1: tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %0 = "xla_hlo.convolution"(%arg0, %arg1) {batch_group_count = 1 : i64, dimension_numbers = {input_batch_dimension = 0 : i64, input_feature_dimension = 3 : i64, input_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>, kernel_input_feature_dimension = 2 : i64, kernel_output_feature_dimension = 3 : i64, kernel_spatial_dimensions = dense<[0, 1]> : tensor<2xi64>, output_batch_dimension = 0 : i64, output_feature_dimension = 3 : i64, output_spatial_dimensions = dense<[1, 2]> : tensor<2xi64>}, feature_group_count = 1 : i64, padding = dense<0> : tensor<2x2xi64>, rhs_dilation = dense<1> : tensor<2xi64>, window_strides = dense<1> : tensor<2xi64>} : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| return %0 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| flow.executable @predict_ex_dispatch_103 attributes {sym_visibility = "private"} { | |
| flow.dispatch.entry @predict_ex_dispatch_103 | |
| module { | |
| func @predict_ex_dispatch_103(%arg0: tensor<1x1x1x2048xf32>, %arg1: tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> { | |
| %cst = constant dense<0.000000e+00> : tensor<f32> | |
| %cst_0 = constant dense<1.000000e+00> : tensor<2048xf32> | |
| %cst_1 = constant dense<0.000000e+00> : tensor<2048xf32> | |
| %cst_2 = constant dense<1.001000e-05> : tensor<f32> | |
| %0 = "xla_hlo.broadcast_in_dim"(%cst_1) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %1 = xla_hlo.add %arg0, %0 : tensor<1x1x1x2048xf32> | |
| %2 = xla_hlo.subtract %1, %0 : tensor<1x1x1x2048xf32> | |
| %3 = "xla_hlo.broadcast_in_dim"(%cst_0) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %4 = xla_hlo.multiply %2, %3 : tensor<1x1x1x2048xf32> | |
| %5 = "xla_hlo.broadcast_in_dim"(%cst_2) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<2048xf32> | |
| %6 = xla_hlo.add %5, %cst_0 : tensor<2048xf32> | |
| %7 = "xla_hlo.sqrt"(%6) : (tensor<2048xf32>) -> tensor<2048xf32> | |
| %8 = "xla_hlo.broadcast_in_dim"(%7) {broadcast_dimensions = dense<3> : tensor<1xi64>} : (tensor<2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %9 = xla_hlo.divide %4, %8 : tensor<1x1x1x2048xf32> | |
| %10 = xla_hlo.add %9, %0 : tensor<1x1x1x2048xf32> | |
| %11 = xla_hlo.add %arg1, %10 : tensor<1x1x1x2048xf32> | |
| %12 = "xla_hlo.broadcast_in_dim"(%cst) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x1x1x2048xf32> | |
| %13 = xla_hlo.maximum %11, %12 : tensor<1x1x1x2048xf32> | |
| return %13 : tensor<1x1x1x2048xf32> | |
| } | |
| } | |
| } | |
| func @predict() -> tensor<1x1x1x2048xf32> attributes {iree.module.export, iree.reflection = {abi = "sip", abiv = 1 : i32, f = "I1!R16!B12!d1d1d1d2048", fv = "1", sip = "I8!S5!k0_0R3!_0"}, tf._input_shapes = ["tfshape$dim { size: 1 } dim { size: 32 } dim { size: 32 } dim { size: 3 }", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true", "tfshape$unknown_rank: true"], tf.signature.is_stateful} { | |
| %cst = constant dense<1.200000e+00> : tensor<1x1x512x2048xf32> | |
| %cst_0 = constant dense<1.200000e+00> : tensor<3x3x512x512xf32> | |
| %cst_1 = constant dense<1.200000e+00> : tensor<1x1x2048x512xf32> | |
| %cst_2 = constant dense<1.200000e+00> : tensor<1x1x1024x512xf32> | |
| %cst_3 = constant dense<1.200000e+00> : tensor<1x1x1024x2048xf32> | |
| %cst_4 = constant dense<1.200000e+00> : tensor<1x1x256x1024xf32> | |
| %cst_5 = constant dense<1.200000e+00> : tensor<3x3x256x256xf32> | |
| %cst_6 = constant dense<1.200000e+00> : tensor<1x1x1024x256xf32> | |
| %cst_7 = constant dense<1.200000e+00> : tensor<1x1x512x256xf32> | |
| %cst_8 = constant dense<1.200000e+00> : tensor<1x1x512x1024xf32> | |
| %cst_9 = constant dense<1.200000e+00> : tensor<1x1x128x512xf32> | |
| %cst_10 = constant dense<1.200000e+00> : tensor<3x3x128x128xf32> | |
| %cst_11 = constant dense<1.200000e+00> : tensor<1x1x512x128xf32> | |
| %cst_12 = constant dense<1.200000e+00> : tensor<1x1x256x128xf32> | |
| %cst_13 = constant dense<1.200000e+00> : tensor<1x1x256x512xf32> | |
| %cst_14 = constant dense<1.200000e+00> : tensor<1x1x64x256xf32> | |
| %cst_15 = constant dense<1.200000e+00> : tensor<3x3x64x64xf32> | |
| %cst_16 = constant dense<1.200000e+00> : tensor<1x1x256x64xf32> | |
| %cst_17 = constant dense<1.200000e+00> : tensor<1x1x64x64xf32> | |
| %cst_18 = constant dense<1.200000e+00> : tensor<7x7x3x64xf32> | |
| %cst_19 = constant dense<1.200000e+00> : tensor<1x32x32x3xf32> | |
| %c512 = constant 512 : index | |
| %c1024 = constant 1024 : index | |
| %c2048 = constant 2048 : index | |
| %c8192 = constant 8192 : index | |
| %c4096 = constant 4096 : index | |
| %c20736 = constant 20736 : index | |
| %c16384 = constant 16384 : index | |
| %0 = iree.do_not_optimize(%cst_19) : tensor<1x32x32x3xf32> | |
| %1 = flow.ex.stream.fragment(%arg0 = %c16384 : index, %arg1 = %0 : tensor<1x32x32x3xf32>, %arg2 = %cst_18 : tensor<7x7x3x64xf32>, %arg3 = %c20736 : index, %arg4 = %c4096 : index, %arg5 = %cst_14 : tensor<1x1x64x256xf32>, %arg6 = %cst_17 : tensor<1x1x64x64xf32>, %arg7 = %cst_15 : tensor<3x3x64x64xf32>, %arg8 = %cst_16 : tensor<1x1x256x64xf32>, %arg9 = %c8192 : index, %arg10 = %cst_13 : tensor<1x1x256x512xf32>, %arg11 = %c2048 : index, %arg12 = %cst_12 : tensor<1x1x256x128xf32>, %arg13 = %cst_10 : tensor<3x3x128x128xf32>, %arg14 = %cst_9 : tensor<1x1x128x512xf32>, %arg15 = %cst_11 : tensor<1x1x512x128xf32>, %arg16 = %cst_8 : tensor<1x1x512x1024xf32>, %arg17 = %c1024 : index, %arg18 = %cst_7 : tensor<1x1x512x256xf32>, %arg19 = %cst_5 : tensor<3x3x256x256xf32>, %arg20 = %cst_4 : tensor<1x1x256x1024xf32>, %arg21 = %cst_6 : tensor<1x1x1024x256xf32>, %arg22 = %cst_3 : tensor<1x1x1024x2048xf32>, %arg23 = %c512 : index, %arg24 = %cst_2 : tensor<1x1x1024x512xf32>, %arg25 = %cst_0 : tensor<3x3x512x512xf32>, %arg26 = %cst : tensor<1x1x512x2048xf32>, %arg27 = %cst_1 : tensor<1x1x2048x512xf32>) -> tensor<1x1x1x2048xf32> { | |
| %2 = flow.dispatch @predict_ex_dispatch_0::@predict_ex_dispatch_0[%arg0 : index](%arg1, %arg2) : (tensor<1x32x32x3xf32>, tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32> | |
| %3:4 = flow.dispatch @predict_ex_dispatch_1::@predict_ex_dispatch_1[%arg0 : index](%2) : (tensor<1x16x16x64xf32>) -> (tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) | |
| %4 = flow.dispatch @predict_ex_dispatch_2::@predict_ex_dispatch_2[%arg3 : index](%3#0) : (tensor<1x16x16x64xf32>) -> tensor<1x18x18x64xf32> | |
| %5 = flow.dispatch @predict_ex_dispatch_3::@predict_ex_dispatch_3[%arg4 : index](%4) : (tensor<1x18x18x64xf32>) -> tensor<1x8x8x64xf32> | |
| %6 = flow.dispatch @predict_ex_dispatch_4::@predict_ex_dispatch_4[%arg0 : index](%5, %arg5) : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %7 = flow.dispatch @predict_ex_dispatch_5::@predict_ex_dispatch_5[%arg4 : index](%5, %arg6) : (tensor<1x8x8x64xf32>, tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %8:7 = flow.dispatch @predict_ex_dispatch_6::@predict_ex_dispatch_6[%arg4 : index](%7) : (tensor<1x8x8x64xf32>) -> (tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) | |
| %9 = flow.dispatch @predict_ex_dispatch_7::@predict_ex_dispatch_7[%arg4 : index](%8#0, %arg7) : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %10 = flow.dispatch @predict_ex_dispatch_8::@predict_ex_dispatch_8[%arg4 : index](%9) : (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> | |
| %11 = flow.dispatch @predict_ex_dispatch_9::@predict_ex_dispatch_9[%arg0 : index](%10, %arg5) : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %12 = flow.dispatch @predict_ex_dispatch_10::@predict_ex_dispatch_10[%arg0 : index](%6, %11, %3#1) : (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> | |
| %13 = flow.dispatch @predict_ex_dispatch_11::@predict_ex_dispatch_11[%arg4 : index](%12, %arg8) : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| %14 = flow.dispatch @predict_ex_dispatch_12::@predict_ex_dispatch_12[%arg4 : index](%13) : (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> | |
| %15 = flow.dispatch @predict_ex_dispatch_13::@predict_ex_dispatch_13[%arg4 : index](%14, %arg7) : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %16 = flow.dispatch @predict_ex_dispatch_14::@predict_ex_dispatch_14[%arg4 : index](%15) : (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> | |
| %17 = flow.dispatch @predict_ex_dispatch_15::@predict_ex_dispatch_15[%arg0 : index](%16, %arg5) : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %18 = flow.dispatch @predict_ex_dispatch_16::@predict_ex_dispatch_16[%arg0 : index](%17, %12, %3#2) : (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> | |
| %19 = flow.dispatch @predict_ex_dispatch_17::@predict_ex_dispatch_17[%arg4 : index](%18, %arg8) : (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32> | |
| %20 = flow.dispatch @predict_ex_dispatch_18::@predict_ex_dispatch_18[%arg4 : index](%19) : (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> | |
| %21 = flow.dispatch @predict_ex_dispatch_19::@predict_ex_dispatch_19[%arg4 : index](%20, %arg7) : (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32> | |
| %22 = flow.dispatch @predict_ex_dispatch_20::@predict_ex_dispatch_20[%arg4 : index](%21) : (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32> | |
| %23 = flow.dispatch @predict_ex_dispatch_21::@predict_ex_dispatch_21[%arg0 : index](%22, %arg5) : (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32> | |
| %24 = flow.dispatch @predict_ex_dispatch_22::@predict_ex_dispatch_22[%arg0 : index](%23, %18, %3#3) : (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32> | |
| %25 = flow.dispatch @predict_ex_dispatch_23::@predict_ex_dispatch_23[%arg9 : index](%24, %arg10) : (tensor<1x8x8x256xf32>, tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32> | |
| %26 = flow.dispatch @predict_ex_dispatch_24::@predict_ex_dispatch_24[%arg11 : index](%24, %arg12) : (tensor<1x8x8x256xf32>, tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32> | |
| %27:3 = flow.dispatch @predict_ex_dispatch_25::@predict_ex_dispatch_25[%arg11 : index](%26) : (tensor<1x4x4x128xf32>) -> (tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) | |
| %28 = flow.dispatch @predict_ex_dispatch_26::@predict_ex_dispatch_26[%arg11 : index](%27#0, %arg13) : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %29 = flow.dispatch @predict_ex_dispatch_27::@predict_ex_dispatch_27[%arg11 : index](%28) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %30 = flow.dispatch @predict_ex_dispatch_28::@predict_ex_dispatch_28[%arg9 : index](%29, %arg14) : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %31:4 = flow.dispatch @predict_ex_dispatch_29::@predict_ex_dispatch_29[%arg9 : index](%25, %30) : (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) | |
| %32 = flow.dispatch @predict_ex_dispatch_30::@predict_ex_dispatch_30[%arg11 : index](%31#0, %arg15) : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %33 = flow.dispatch @predict_ex_dispatch_31::@predict_ex_dispatch_31[%arg11 : index](%32) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %34 = flow.dispatch @predict_ex_dispatch_32::@predict_ex_dispatch_32[%arg11 : index](%33, %arg13) : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %35 = flow.dispatch @predict_ex_dispatch_33::@predict_ex_dispatch_33[%arg11 : index](%34) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %36 = flow.dispatch @predict_ex_dispatch_34::@predict_ex_dispatch_34[%arg9 : index](%35, %arg14) : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %37 = flow.dispatch @predict_ex_dispatch_35::@predict_ex_dispatch_35[%arg9 : index](%36, %31#0, %31#1) : (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> | |
| %38 = flow.dispatch @predict_ex_dispatch_36::@predict_ex_dispatch_36[%arg11 : index](%37, %arg15) : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %39 = flow.dispatch @predict_ex_dispatch_37::@predict_ex_dispatch_37[%arg11 : index](%38) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %40 = flow.dispatch @predict_ex_dispatch_38::@predict_ex_dispatch_38[%arg11 : index](%39, %arg13) : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %41 = flow.dispatch @predict_ex_dispatch_39::@predict_ex_dispatch_39[%arg11 : index](%40) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %42 = flow.dispatch @predict_ex_dispatch_40::@predict_ex_dispatch_40[%arg9 : index](%41, %arg14) : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %43 = flow.dispatch @predict_ex_dispatch_41::@predict_ex_dispatch_41[%arg9 : index](%42, %37, %31#2) : (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> | |
| %44 = flow.dispatch @predict_ex_dispatch_42::@predict_ex_dispatch_42[%arg11 : index](%43, %arg15) : (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32> | |
| %45 = flow.dispatch @predict_ex_dispatch_43::@predict_ex_dispatch_43[%arg11 : index](%44) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %46 = flow.dispatch @predict_ex_dispatch_44::@predict_ex_dispatch_44[%arg11 : index](%45, %arg13) : (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32> | |
| %47 = flow.dispatch @predict_ex_dispatch_45::@predict_ex_dispatch_45[%arg11 : index](%46) : (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32> | |
| %48 = flow.dispatch @predict_ex_dispatch_46::@predict_ex_dispatch_46[%arg9 : index](%47, %arg14) : (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32> | |
| %49 = flow.dispatch @predict_ex_dispatch_47::@predict_ex_dispatch_47[%arg9 : index](%48, %43, %31#3) : (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32> | |
| %50 = flow.dispatch @predict_ex_dispatch_48::@predict_ex_dispatch_48[%arg4 : index](%49, %arg16) : (tensor<1x4x4x512xf32>, tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %51 = flow.dispatch @predict_ex_dispatch_49::@predict_ex_dispatch_49[%arg17 : index](%49, %arg18) : (tensor<1x4x4x512xf32>, tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32> | |
| %52 = flow.dispatch @predict_ex_dispatch_50::@predict_ex_dispatch_50[%arg17 : index](%51) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %53 = flow.dispatch @predict_ex_dispatch_51::@predict_ex_dispatch_51[%arg17 : index](%52, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %54 = flow.dispatch @predict_ex_dispatch_52::@predict_ex_dispatch_52[%arg17 : index](%53) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %55 = flow.dispatch @predict_ex_dispatch_53::@predict_ex_dispatch_53[%arg4 : index](%54, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %56 = flow.dispatch @predict_ex_dispatch_54::@predict_ex_dispatch_54[%arg4 : index](%50, %55, %8#1) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %57 = flow.dispatch @predict_ex_dispatch_55::@predict_ex_dispatch_55[%arg17 : index](%56, %arg21) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %58 = flow.dispatch @predict_ex_dispatch_56::@predict_ex_dispatch_56[%arg17 : index](%57) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %59 = flow.dispatch @predict_ex_dispatch_57::@predict_ex_dispatch_57[%arg17 : index](%58, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %60 = flow.dispatch @predict_ex_dispatch_58::@predict_ex_dispatch_58[%arg17 : index](%59) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %61 = flow.dispatch @predict_ex_dispatch_59::@predict_ex_dispatch_59[%arg4 : index](%60, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %62 = flow.dispatch @predict_ex_dispatch_60::@predict_ex_dispatch_60[%arg4 : index](%61, %56, %8#2) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %63 = flow.dispatch @predict_ex_dispatch_61::@predict_ex_dispatch_61[%arg17 : index](%62, %arg21) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %64 = flow.dispatch @predict_ex_dispatch_62::@predict_ex_dispatch_62[%arg17 : index](%63) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %65 = flow.dispatch @predict_ex_dispatch_63::@predict_ex_dispatch_63[%arg17 : index](%64, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %66 = flow.dispatch @predict_ex_dispatch_64::@predict_ex_dispatch_64[%arg17 : index](%65) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %67 = flow.dispatch @predict_ex_dispatch_65::@predict_ex_dispatch_65[%arg4 : index](%66, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %68 = flow.dispatch @predict_ex_dispatch_66::@predict_ex_dispatch_66[%arg4 : index](%67, %62, %8#3) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %69 = flow.dispatch @predict_ex_dispatch_67::@predict_ex_dispatch_67[%arg17 : index](%68, %arg21) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %70 = flow.dispatch @predict_ex_dispatch_68::@predict_ex_dispatch_68[%arg17 : index](%69) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %71 = flow.dispatch @predict_ex_dispatch_69::@predict_ex_dispatch_69[%arg17 : index](%70, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %72 = flow.dispatch @predict_ex_dispatch_70::@predict_ex_dispatch_70[%arg17 : index](%71) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %73 = flow.dispatch @predict_ex_dispatch_71::@predict_ex_dispatch_71[%arg4 : index](%72, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %74 = flow.dispatch @predict_ex_dispatch_72::@predict_ex_dispatch_72[%arg4 : index](%73, %68, %8#4) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %75 = flow.dispatch @predict_ex_dispatch_73::@predict_ex_dispatch_73[%arg17 : index](%74, %arg21) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %76 = flow.dispatch @predict_ex_dispatch_74::@predict_ex_dispatch_74[%arg17 : index](%75) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %77 = flow.dispatch @predict_ex_dispatch_75::@predict_ex_dispatch_75[%arg17 : index](%76, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %78 = flow.dispatch @predict_ex_dispatch_76::@predict_ex_dispatch_76[%arg17 : index](%77) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %79 = flow.dispatch @predict_ex_dispatch_77::@predict_ex_dispatch_77[%arg4 : index](%78, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %80 = flow.dispatch @predict_ex_dispatch_78::@predict_ex_dispatch_78[%arg4 : index](%79, %74, %8#5) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %81 = flow.dispatch @predict_ex_dispatch_79::@predict_ex_dispatch_79[%arg17 : index](%80, %arg21) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32> | |
| %82 = flow.dispatch @predict_ex_dispatch_80::@predict_ex_dispatch_80[%arg17 : index](%81) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %83 = flow.dispatch @predict_ex_dispatch_81::@predict_ex_dispatch_81[%arg17 : index](%82, %arg19) : (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32> | |
| %84 = flow.dispatch @predict_ex_dispatch_82::@predict_ex_dispatch_82[%arg17 : index](%83) : (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32> | |
| %85 = flow.dispatch @predict_ex_dispatch_83::@predict_ex_dispatch_83[%arg4 : index](%84, %arg20) : (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %86 = flow.dispatch @predict_ex_dispatch_84::@predict_ex_dispatch_84[%arg4 : index](%85, %80, %8#6) : (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32> | |
| %87 = flow.dispatch @predict_ex_dispatch_85::@predict_ex_dispatch_85[%arg11 : index](%86, %arg22) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %88 = flow.dispatch @predict_ex_dispatch_86::@predict_ex_dispatch_86[%arg23 : index](%86, %arg24) : (tensor<1x2x2x1024xf32>, tensor<1x1x1024x512xf32>) -> tensor<1x1x1x512xf32> | |
| %89 = flow.dispatch @predict_ex_dispatch_87::@predict_ex_dispatch_87[%arg23 : index](%88) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %90 = flow.dispatch @predict_ex_dispatch_88::@predict_ex_dispatch_88[%arg23 : index](%89, %arg25) : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %91 = flow.dispatch @predict_ex_dispatch_89::@predict_ex_dispatch_89[%arg23 : index](%90) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %92 = flow.dispatch @predict_ex_dispatch_90::@predict_ex_dispatch_90[%arg11 : index](%91, %arg26) : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %93 = flow.dispatch @predict_ex_dispatch_91::@predict_ex_dispatch_91[%arg11 : index](%87, %92, %27#1) : (tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %94 = flow.dispatch @predict_ex_dispatch_92::@predict_ex_dispatch_92[%arg23 : index](%93, %arg27) : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| %95 = flow.dispatch @predict_ex_dispatch_93::@predict_ex_dispatch_93[%arg23 : index](%94) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %96 = flow.dispatch @predict_ex_dispatch_94::@predict_ex_dispatch_94[%arg23 : index](%95, %arg25) : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %97 = flow.dispatch @predict_ex_dispatch_95::@predict_ex_dispatch_95[%arg23 : index](%96) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %98 = flow.dispatch @predict_ex_dispatch_96::@predict_ex_dispatch_96[%arg11 : index](%97, %arg26) : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %99 = flow.dispatch @predict_ex_dispatch_97::@predict_ex_dispatch_97[%arg11 : index](%98, %93, %27#2) : (tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %100 = flow.dispatch @predict_ex_dispatch_98::@predict_ex_dispatch_98[%arg23 : index](%99, %arg27) : (tensor<1x1x1x2048xf32>, tensor<1x1x2048x512xf32>) -> tensor<1x1x1x512xf32> | |
| %101 = flow.dispatch @predict_ex_dispatch_99::@predict_ex_dispatch_99[%arg23 : index](%100) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %102 = flow.dispatch @predict_ex_dispatch_100::@predict_ex_dispatch_100[%arg23 : index](%101, %arg25) : (tensor<1x1x1x512xf32>, tensor<3x3x512x512xf32>) -> tensor<1x1x1x512xf32> | |
| %103 = flow.dispatch @predict_ex_dispatch_101::@predict_ex_dispatch_101[%arg23 : index](%102) : (tensor<1x1x1x512xf32>) -> tensor<1x1x1x512xf32> | |
| %104 = flow.dispatch @predict_ex_dispatch_102::@predict_ex_dispatch_102[%arg11 : index](%103, %arg26) : (tensor<1x1x1x512xf32>, tensor<1x1x512x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| %105 = flow.dispatch @predict_ex_dispatch_103::@predict_ex_dispatch_103[%arg11 : index](%104, %99) : (tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>) -> tensor<1x1x1x2048xf32> | |
| flow.return %105 : tensor<1x1x1x2048xf32> | |
| } | |
| return %1 : tensor<1x1x1x2048xf32> | |
| } | |
| } |
This file has been truncated, but you can view the full file.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 370 : i32}} { | |
| func @_device_match_id_0_initializer() -> i1 attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %0 = hal.device.match.id %dev, pattern = ["vmla"] : (!hal.device) -> i1 | |
| return %0 : i1 | |
| } | |
| hal.variable @_device_match_id_0 init(@_device_match_id_0_initializer) : i1 attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_0 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_1 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_2 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_3 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_4 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_5 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_6 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_7 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_8 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_9 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_10 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_11 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_12 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_13 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_14 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_15 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_16 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_17 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_18 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_19 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_20 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_21 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_22 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_23 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_24 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_25 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_26 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_27 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_28 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_29 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_30 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_31 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_32 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_33 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_34 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_35 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_36 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_37 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_38 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_39 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_40 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_41 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_42 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_43 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_44 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_45 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_46 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_47 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_48 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_49 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_50 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_51 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_52 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_53 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_54 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_55 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_56 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_57 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_58 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_59 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_60 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_61 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_62 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_63 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_64 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_65 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_66 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_67 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_68 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_69 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_70 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_71 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_72 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_73 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_74 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_75 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_76 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_77 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_78 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_79 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_80 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_81 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_82 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_83 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_84 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_85 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_86 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_87 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_88 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_89 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_90 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_91 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_92 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_93 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_94 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_95 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_96 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_97 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_98 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_99 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_100 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_101 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_102 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_executable_predict_ex_dispatch_103 mutable : !hal.executable attributes {sym_visibility = "private"} | |
| hal.variable @_descriptor_set_layout_0 init(@_descriptor_set_layout_0_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_0_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_0 init(@_executable_layout_0_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_0_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_0 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_1 init(@_descriptor_set_layout_1_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_1_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<3, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<4, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_1 init(@_executable_layout_1_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_1_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_1 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_2 init(@_descriptor_set_layout_2_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_2_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_2 init(@_executable_layout_2_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_2_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_2 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_3 init(@_descriptor_set_layout_3_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_3_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<3, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<4, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<5, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<6, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<7, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_3 init(@_executable_layout_3_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_3_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_3 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_4 init(@_descriptor_set_layout_4_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_4_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<3, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_4 init(@_executable_layout_4_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_4_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_4 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_5 init(@_descriptor_set_layout_5_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_5_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<3, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_5 init(@_executable_layout_5_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_5_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_5 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_descriptor_set_layout_6 init(@_descriptor_set_layout_6_initializer) : !hal.descriptor_set_layout attributes {sym_visibility = "private"} | |
| func @_descriptor_set_layout_6_initializer() -> !hal.descriptor_set_layout attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %descriptor_set_layout = hal.descriptor_set_layout.create %dev, "PushOnly", bindings = [#hal.descriptor_set_layout_binding<0, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<1, "StorageBuffer", "Read">, #hal.descriptor_set_layout_binding<2, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<3, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<4, "StorageBuffer", "Write|Discard">, #hal.descriptor_set_layout_binding<5, "StorageBuffer", "Write|Discard">] : !hal.descriptor_set_layout | |
| return %descriptor_set_layout : !hal.descriptor_set_layout | |
| } | |
| hal.variable @_executable_layout_6 init(@_executable_layout_6_initializer) : !hal.executable_layout attributes {sym_visibility = "private"} | |
| func @_executable_layout_6_initializer() -> !hal.executable_layout attributes {sym_visibility = "private"} { | |
| %0 = hal.variable.load @_descriptor_set_layout_6 : !hal.descriptor_set_layout | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_layout = hal.executable_layout.create %dev, set_layouts = [%0], push_constants = 0 : !hal.executable_layout | |
| return %executable_layout : !hal.executable_layout | |
| } | |
| hal.variable @_executable_cache init(@_executable_cache_initializer) : !hal.executable_cache | |
| func @_executable_cache_initializer() -> !hal.executable_cache attributes {sym_visibility = "private"} { | |
| %dev = hal.ex.shared_device : !hal.device | |
| %executable_cache_default = hal.executable_cache.create %dev, identifier = "default" : !hal.executable_cache | |
| %0 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_0 = hal.executable_cache.prepare %executable_cache_default, layout = %0, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_0 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_0, @_executable_predict_ex_dispatch_0 : !hal.executable | |
| %1 = hal.variable.load @_executable_layout_1 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_1 = hal.executable_cache.prepare %executable_cache_default, layout = %1, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_1 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_1, @_executable_predict_ex_dispatch_1 : !hal.executable | |
| %2 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_2 = hal.executable_cache.prepare %executable_cache_default, layout = %2, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_2 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_2, @_executable_predict_ex_dispatch_2 : !hal.executable | |
| %3 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_3 = hal.executable_cache.prepare %executable_cache_default, layout = %3, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_3 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_3, @_executable_predict_ex_dispatch_3 : !hal.executable | |
| %4 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_4 = hal.executable_cache.prepare %executable_cache_default, layout = %4, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_4 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_4, @_executable_predict_ex_dispatch_4 : !hal.executable | |
| %5 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_5 = hal.executable_cache.prepare %executable_cache_default, layout = %5, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_5 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_5, @_executable_predict_ex_dispatch_5 : !hal.executable | |
| %6 = hal.variable.load @_executable_layout_3 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_6 = hal.executable_cache.prepare %executable_cache_default, layout = %6, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_6 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_6, @_executable_predict_ex_dispatch_6 : !hal.executable | |
| %7 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_7 = hal.executable_cache.prepare %executable_cache_default, layout = %7, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_7 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_7, @_executable_predict_ex_dispatch_7 : !hal.executable | |
| %8 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_8 = hal.executable_cache.prepare %executable_cache_default, layout = %8, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_8 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_8, @_executable_predict_ex_dispatch_8 : !hal.executable | |
| %9 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_9 = hal.executable_cache.prepare %executable_cache_default, layout = %9, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_9 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_9, @_executable_predict_ex_dispatch_9 : !hal.executable | |
| %10 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_10 = hal.executable_cache.prepare %executable_cache_default, layout = %10, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_10 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_10, @_executable_predict_ex_dispatch_10 : !hal.executable | |
| %11 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_11 = hal.executable_cache.prepare %executable_cache_default, layout = %11, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_11 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_11, @_executable_predict_ex_dispatch_11 : !hal.executable | |
| %12 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_12 = hal.executable_cache.prepare %executable_cache_default, layout = %12, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_12 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_12, @_executable_predict_ex_dispatch_12 : !hal.executable | |
| %13 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_13 = hal.executable_cache.prepare %executable_cache_default, layout = %13, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_13 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_13, @_executable_predict_ex_dispatch_13 : !hal.executable | |
| %14 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_14 = hal.executable_cache.prepare %executable_cache_default, layout = %14, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_14 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_14, @_executable_predict_ex_dispatch_14 : !hal.executable | |
| %15 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_15 = hal.executable_cache.prepare %executable_cache_default, layout = %15, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_15 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_15, @_executable_predict_ex_dispatch_15 : !hal.executable | |
| %16 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_16 = hal.executable_cache.prepare %executable_cache_default, layout = %16, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_16 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_16, @_executable_predict_ex_dispatch_16 : !hal.executable | |
| %17 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_17 = hal.executable_cache.prepare %executable_cache_default, layout = %17, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_17 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_17, @_executable_predict_ex_dispatch_17 : !hal.executable | |
| %18 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_18 = hal.executable_cache.prepare %executable_cache_default, layout = %18, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_18 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_18, @_executable_predict_ex_dispatch_18 : !hal.executable | |
| %19 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_19 = hal.executable_cache.prepare %executable_cache_default, layout = %19, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_19 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_19, @_executable_predict_ex_dispatch_19 : !hal.executable | |
| %20 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_20 = hal.executable_cache.prepare %executable_cache_default, layout = %20, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_20 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_20, @_executable_predict_ex_dispatch_20 : !hal.executable | |
| %21 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_21 = hal.executable_cache.prepare %executable_cache_default, layout = %21, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_21 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_21, @_executable_predict_ex_dispatch_21 : !hal.executable | |
| %22 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_22 = hal.executable_cache.prepare %executable_cache_default, layout = %22, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_22 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_22, @_executable_predict_ex_dispatch_22 : !hal.executable | |
| %23 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_23 = hal.executable_cache.prepare %executable_cache_default, layout = %23, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_23 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_23, @_executable_predict_ex_dispatch_23 : !hal.executable | |
| %24 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_24 = hal.executable_cache.prepare %executable_cache_default, layout = %24, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_24 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_24, @_executable_predict_ex_dispatch_24 : !hal.executable | |
| %25 = hal.variable.load @_executable_layout_5 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_25 = hal.executable_cache.prepare %executable_cache_default, layout = %25, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_25 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_25, @_executable_predict_ex_dispatch_25 : !hal.executable | |
| %26 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_26 = hal.executable_cache.prepare %executable_cache_default, layout = %26, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_26 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_26, @_executable_predict_ex_dispatch_26 : !hal.executable | |
| %27 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_27 = hal.executable_cache.prepare %executable_cache_default, layout = %27, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_27 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_27, @_executable_predict_ex_dispatch_27 : !hal.executable | |
| %28 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_28 = hal.executable_cache.prepare %executable_cache_default, layout = %28, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_28 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_28, @_executable_predict_ex_dispatch_28 : !hal.executable | |
| %29 = hal.variable.load @_executable_layout_6 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_29 = hal.executable_cache.prepare %executable_cache_default, layout = %29, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_29 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_29, @_executable_predict_ex_dispatch_29 : !hal.executable | |
| %30 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_30 = hal.executable_cache.prepare %executable_cache_default, layout = %30, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_30 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_30, @_executable_predict_ex_dispatch_30 : !hal.executable | |
| %31 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_31 = hal.executable_cache.prepare %executable_cache_default, layout = %31, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_31 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_31, @_executable_predict_ex_dispatch_31 : !hal.executable | |
| %32 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_32 = hal.executable_cache.prepare %executable_cache_default, layout = %32, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_32 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_32, @_executable_predict_ex_dispatch_32 : !hal.executable | |
| %33 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_33 = hal.executable_cache.prepare %executable_cache_default, layout = %33, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_33 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_33, @_executable_predict_ex_dispatch_33 : !hal.executable | |
| %34 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_34 = hal.executable_cache.prepare %executable_cache_default, layout = %34, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_34 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_34, @_executable_predict_ex_dispatch_34 : !hal.executable | |
| %35 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_35 = hal.executable_cache.prepare %executable_cache_default, layout = %35, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_35 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_35, @_executable_predict_ex_dispatch_35 : !hal.executable | |
| %36 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_36 = hal.executable_cache.prepare %executable_cache_default, layout = %36, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_36 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_36, @_executable_predict_ex_dispatch_36 : !hal.executable | |
| %37 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_37 = hal.executable_cache.prepare %executable_cache_default, layout = %37, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_37 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_37, @_executable_predict_ex_dispatch_37 : !hal.executable | |
| %38 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_38 = hal.executable_cache.prepare %executable_cache_default, layout = %38, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_38 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_38, @_executable_predict_ex_dispatch_38 : !hal.executable | |
| %39 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_39 = hal.executable_cache.prepare %executable_cache_default, layout = %39, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_39 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_39, @_executable_predict_ex_dispatch_39 : !hal.executable | |
| %40 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_40 = hal.executable_cache.prepare %executable_cache_default, layout = %40, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_40 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_40, @_executable_predict_ex_dispatch_40 : !hal.executable | |
| %41 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_41 = hal.executable_cache.prepare %executable_cache_default, layout = %41, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_41 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_41, @_executable_predict_ex_dispatch_41 : !hal.executable | |
| %42 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_42 = hal.executable_cache.prepare %executable_cache_default, layout = %42, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_42 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_42, @_executable_predict_ex_dispatch_42 : !hal.executable | |
| %43 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_43 = hal.executable_cache.prepare %executable_cache_default, layout = %43, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_43 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_43, @_executable_predict_ex_dispatch_43 : !hal.executable | |
| %44 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_44 = hal.executable_cache.prepare %executable_cache_default, layout = %44, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_44 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_44, @_executable_predict_ex_dispatch_44 : !hal.executable | |
| %45 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_45 = hal.executable_cache.prepare %executable_cache_default, layout = %45, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_45 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_45, @_executable_predict_ex_dispatch_45 : !hal.executable | |
| %46 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_46 = hal.executable_cache.prepare %executable_cache_default, layout = %46, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_46 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_46, @_executable_predict_ex_dispatch_46 : !hal.executable | |
| %47 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_47 = hal.executable_cache.prepare %executable_cache_default, layout = %47, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_47 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_47, @_executable_predict_ex_dispatch_47 : !hal.executable | |
| %48 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_48 = hal.executable_cache.prepare %executable_cache_default, layout = %48, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_48 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_48, @_executable_predict_ex_dispatch_48 : !hal.executable | |
| %49 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_49 = hal.executable_cache.prepare %executable_cache_default, layout = %49, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_49 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_49, @_executable_predict_ex_dispatch_49 : !hal.executable | |
| %50 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_50 = hal.executable_cache.prepare %executable_cache_default, layout = %50, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_50 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_50, @_executable_predict_ex_dispatch_50 : !hal.executable | |
| %51 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_51 = hal.executable_cache.prepare %executable_cache_default, layout = %51, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_51 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_51, @_executable_predict_ex_dispatch_51 : !hal.executable | |
| %52 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_52 = hal.executable_cache.prepare %executable_cache_default, layout = %52, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_52 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_52, @_executable_predict_ex_dispatch_52 : !hal.executable | |
| %53 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_53 = hal.executable_cache.prepare %executable_cache_default, layout = %53, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_53 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_53, @_executable_predict_ex_dispatch_53 : !hal.executable | |
| %54 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_54 = hal.executable_cache.prepare %executable_cache_default, layout = %54, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_54 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_54, @_executable_predict_ex_dispatch_54 : !hal.executable | |
| %55 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_55 = hal.executable_cache.prepare %executable_cache_default, layout = %55, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_55 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_55, @_executable_predict_ex_dispatch_55 : !hal.executable | |
| %56 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_56 = hal.executable_cache.prepare %executable_cache_default, layout = %56, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_56 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_56, @_executable_predict_ex_dispatch_56 : !hal.executable | |
| %57 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_57 = hal.executable_cache.prepare %executable_cache_default, layout = %57, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_57 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_57, @_executable_predict_ex_dispatch_57 : !hal.executable | |
| %58 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_58 = hal.executable_cache.prepare %executable_cache_default, layout = %58, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_58 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_58, @_executable_predict_ex_dispatch_58 : !hal.executable | |
| %59 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_59 = hal.executable_cache.prepare %executable_cache_default, layout = %59, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_59 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_59, @_executable_predict_ex_dispatch_59 : !hal.executable | |
| %60 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_60 = hal.executable_cache.prepare %executable_cache_default, layout = %60, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_60 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_60, @_executable_predict_ex_dispatch_60 : !hal.executable | |
| %61 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_61 = hal.executable_cache.prepare %executable_cache_default, layout = %61, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_61 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_61, @_executable_predict_ex_dispatch_61 : !hal.executable | |
| %62 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_62 = hal.executable_cache.prepare %executable_cache_default, layout = %62, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_62 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_62, @_executable_predict_ex_dispatch_62 : !hal.executable | |
| %63 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_63 = hal.executable_cache.prepare %executable_cache_default, layout = %63, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_63 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_63, @_executable_predict_ex_dispatch_63 : !hal.executable | |
| %64 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_64 = hal.executable_cache.prepare %executable_cache_default, layout = %64, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_64 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_64, @_executable_predict_ex_dispatch_64 : !hal.executable | |
| %65 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_65 = hal.executable_cache.prepare %executable_cache_default, layout = %65, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_65 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_65, @_executable_predict_ex_dispatch_65 : !hal.executable | |
| %66 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_66 = hal.executable_cache.prepare %executable_cache_default, layout = %66, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_66 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_66, @_executable_predict_ex_dispatch_66 : !hal.executable | |
| %67 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_67 = hal.executable_cache.prepare %executable_cache_default, layout = %67, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_67 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_67, @_executable_predict_ex_dispatch_67 : !hal.executable | |
| %68 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_68 = hal.executable_cache.prepare %executable_cache_default, layout = %68, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_68 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_68, @_executable_predict_ex_dispatch_68 : !hal.executable | |
| %69 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_69 = hal.executable_cache.prepare %executable_cache_default, layout = %69, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_69 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_69, @_executable_predict_ex_dispatch_69 : !hal.executable | |
| %70 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_70 = hal.executable_cache.prepare %executable_cache_default, layout = %70, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_70 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_70, @_executable_predict_ex_dispatch_70 : !hal.executable | |
| %71 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_71 = hal.executable_cache.prepare %executable_cache_default, layout = %71, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_71 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_71, @_executable_predict_ex_dispatch_71 : !hal.executable | |
| %72 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_72 = hal.executable_cache.prepare %executable_cache_default, layout = %72, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_72 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_72, @_executable_predict_ex_dispatch_72 : !hal.executable | |
| %73 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_73 = hal.executable_cache.prepare %executable_cache_default, layout = %73, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_73 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_73, @_executable_predict_ex_dispatch_73 : !hal.executable | |
| %74 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_74 = hal.executable_cache.prepare %executable_cache_default, layout = %74, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_74 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_74, @_executable_predict_ex_dispatch_74 : !hal.executable | |
| %75 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_75 = hal.executable_cache.prepare %executable_cache_default, layout = %75, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_75 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_75, @_executable_predict_ex_dispatch_75 : !hal.executable | |
| %76 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_76 = hal.executable_cache.prepare %executable_cache_default, layout = %76, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_76 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_76, @_executable_predict_ex_dispatch_76 : !hal.executable | |
| %77 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_77 = hal.executable_cache.prepare %executable_cache_default, layout = %77, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_77 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_77, @_executable_predict_ex_dispatch_77 : !hal.executable | |
| %78 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_78 = hal.executable_cache.prepare %executable_cache_default, layout = %78, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_78 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_78, @_executable_predict_ex_dispatch_78 : !hal.executable | |
| %79 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_79 = hal.executable_cache.prepare %executable_cache_default, layout = %79, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_79 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_79, @_executable_predict_ex_dispatch_79 : !hal.executable | |
| %80 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_80 = hal.executable_cache.prepare %executable_cache_default, layout = %80, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_80 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_80, @_executable_predict_ex_dispatch_80 : !hal.executable | |
| %81 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_81 = hal.executable_cache.prepare %executable_cache_default, layout = %81, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_81 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_81, @_executable_predict_ex_dispatch_81 : !hal.executable | |
| %82 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_82 = hal.executable_cache.prepare %executable_cache_default, layout = %82, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_82 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_82, @_executable_predict_ex_dispatch_82 : !hal.executable | |
| %83 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_83 = hal.executable_cache.prepare %executable_cache_default, layout = %83, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_83 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_83, @_executable_predict_ex_dispatch_83 : !hal.executable | |
| %84 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_84 = hal.executable_cache.prepare %executable_cache_default, layout = %84, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_84 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_84, @_executable_predict_ex_dispatch_84 : !hal.executable | |
| %85 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_85 = hal.executable_cache.prepare %executable_cache_default, layout = %85, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_85 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_85, @_executable_predict_ex_dispatch_85 : !hal.executable | |
| %86 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_86 = hal.executable_cache.prepare %executable_cache_default, layout = %86, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_86 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_86, @_executable_predict_ex_dispatch_86 : !hal.executable | |
| %87 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_87 = hal.executable_cache.prepare %executable_cache_default, layout = %87, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_87 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_87, @_executable_predict_ex_dispatch_87 : !hal.executable | |
| %88 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_88 = hal.executable_cache.prepare %executable_cache_default, layout = %88, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_88 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_88, @_executable_predict_ex_dispatch_88 : !hal.executable | |
| %89 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_89 = hal.executable_cache.prepare %executable_cache_default, layout = %89, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_89 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_89, @_executable_predict_ex_dispatch_89 : !hal.executable | |
| %90 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_90 = hal.executable_cache.prepare %executable_cache_default, layout = %90, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_90 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_90, @_executable_predict_ex_dispatch_90 : !hal.executable | |
| %91 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_91 = hal.executable_cache.prepare %executable_cache_default, layout = %91, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_91 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_91, @_executable_predict_ex_dispatch_91 : !hal.executable | |
| %92 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_92 = hal.executable_cache.prepare %executable_cache_default, layout = %92, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_92 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_92, @_executable_predict_ex_dispatch_92 : !hal.executable | |
| %93 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_93 = hal.executable_cache.prepare %executable_cache_default, layout = %93, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_93 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_93, @_executable_predict_ex_dispatch_93 : !hal.executable | |
| %94 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_94 = hal.executable_cache.prepare %executable_cache_default, layout = %94, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_94 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_94, @_executable_predict_ex_dispatch_94 : !hal.executable | |
| %95 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_95 = hal.executable_cache.prepare %executable_cache_default, layout = %95, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_95 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_95, @_executable_predict_ex_dispatch_95 : !hal.executable | |
| %96 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_96 = hal.executable_cache.prepare %executable_cache_default, layout = %96, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_96 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_96, @_executable_predict_ex_dispatch_96 : !hal.executable | |
| %97 = hal.variable.load @_executable_layout_4 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_97 = hal.executable_cache.prepare %executable_cache_default, layout = %97, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_97 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_97, @_executable_predict_ex_dispatch_97 : !hal.executable | |
| %98 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_98 = hal.executable_cache.prepare %executable_cache_default, layout = %98, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_98 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_98, @_executable_predict_ex_dispatch_98 : !hal.executable | |
| %99 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_99 = hal.executable_cache.prepare %executable_cache_default, layout = %99, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_99 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_99, @_executable_predict_ex_dispatch_99 : !hal.executable | |
| %100 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_100 = hal.executable_cache.prepare %executable_cache_default, layout = %100, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_100 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_100, @_executable_predict_ex_dispatch_100 : !hal.executable | |
| %101 = hal.variable.load @_executable_layout_2 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_101 = hal.executable_cache.prepare %executable_cache_default, layout = %101, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_101 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_101, @_executable_predict_ex_dispatch_101 : !hal.executable | |
| %102 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_102 = hal.executable_cache.prepare %executable_cache_default, layout = %102, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_102 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_102, @_executable_predict_ex_dispatch_102 : !hal.executable | |
| %103 = hal.variable.load @_executable_layout_0 : !hal.executable_layout | |
| %executable_predict_ex_dispatch_103 = hal.executable_cache.prepare %executable_cache_default, layout = %103, caching_mode = "AliasProvidedData|AllowPersistentCaching|AllowOptimization", @predict_ex_dispatch_103 : !hal.executable | |
| hal.variable.store %executable_predict_ex_dispatch_103, @_executable_predict_ex_dispatch_103 : !hal.executable | |
| return %executable_cache_default : !hal.executable_cache | |
| } | |
| hal.executable @predict_ex_dispatch_0 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_0 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x32x32x3xf32>, tensor<7x7x3x64xf32>) -> tensor<1x16x16x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1040xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_1 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret1, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret2, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret3, set=0, binding=4, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_1 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x16x16x64xf32>) -> (tensor<1x16x16x64xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>)} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000680A000020000000424D4F441800240020001C001800140010000C00000000000800040018000000E8040000980000008000000074000000680000002C0000001400000004000000060000006D6F64756C65000004000000700400004804000028040000080400000D000000AC0300006803000024030000D8020000900200004C0200000C020000D001000094010000580100001C010000E4000000A80000000100000068000000010000002800000004000000540400003C040000240400001804000001000000000000003F0300000A000800B0FCFFFF08000000140000001EFCFFFF04000000010000000300000015000000707265646963745F65785F64697370617463685F31000000E8FCFFFF080000001400000056FCFFFF04000000010000000300000015000000707265646963745F65785F64697370617463685F3100000020FDFFFF080000001C0000008EFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6D61782E6633320000000058FDFFFF0800000018000000C6FCFFFF040000000200000002000000020000000D000000766D6C612E737172742E6633320000008CFDFFFF080000001C000000FAFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6469762E66333200000000C4FDFFFF080000001C00000032FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6D756C2E66333200000000FCFDFFFF080000001C0000006AFDFFFF04000000030000000200000002000000020000000C000000766D6C612E7375622E6633320000000034FEFFFF080000001C000000A2FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6164642E663332000000006CFEFFFF0800000020000000DAFDFFFF0400000004000000020000000000000002000000000000000D000000766D6C612E74696C652E783332000000A8FEFFFF080000002000000016FEFFFF04000000040000000200000000000000020000000000000012000000766D6C612E62726F6164636173742E7833320000E8FEFFFF080000002400000056FEFFFF0400000005000000020000000000000002000000000000000000000010000000766D6C612E6275666665722E636F7079000000002CFFFFFF080000002800000038FFFFFF080000000C00000001000000020000000300000002000000000000000000000010000000766D6C612E6275666665722E766965770000000074FFFFFF080000002000000080FFFFFF080000000C0000000100000002000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000002000000010000000100000011000000766D6C612E6275666665722E636F6E7374000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000020000000300000003000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000AAFFFFFF040000000F00000021766D6C612E696E7465726661636500C6FFFFFF040000000C00000021766D6C612E62756666657200000000E2FFFFFF040000001100000021697265652E627974655F6275666665720006000800040006000000040000000300000069333200DAFFFFFF78030000F6FFFFFF7803000000000A000A000000000004000A0000006804000000000A0008000000000004000A000000580500003F0300000800000900000100010009000100000200090100000003000940000000040009100000000500090800000006000902000000070009030000000800090400000009000B0000000001805201000080010001C0010001800B0100000002805201000080010002C0010002800B0200000003805201000080010003C0010003800B0300000004805201000080010004C00100048052000000800300008000000000010005805203000080030005C000000100010005805202000080010001000100068053060000800400FFFF0400FFFF04000A0003C00300030003000400068003000500050004000000520200008001000100010003805207000080030005C0068003800000520200008001000100010005805208000080030003C00680058000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002800300030003000400038003000500050004000000520200008001000100010007805209000080030005C003C0078000005202000080010002000100038053050000800400FFFF0000FFFF0100030004C0038004000000520200008001000200010004805207000080030003C002C00480000052020000800100020001000280520B000080020004C0028000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002C0030003000300040003800300050005000400000052020000800100010001000280520A000080030007C003C002800000520200008001000100010003805207000080030002C006C0038000005202000080010001000100028053050000800400FFFF0000FFFF04000600018002800300050005000400000052020000800100010001000480520C000080030003C002C0048000005202000080010001000100028053050000800400FFFF0000FFFF0400060001C002800300060006000200000052000000800300008000000300010001805204000080050004C0000001C00000010000005200000080030000800000070001000180520400008005000280000001C00000010000005200000080030000800000080001000180520400008005000280000001C00000010000005200000080030000C000000900010000805204000080050002C0000000C000000100000054000000040000009FF027370001000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000100000000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0400000000000000"> : vector<2692xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_2 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_2 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x16x16x64xf32>) -> tensor<1x18x18x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1108xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_3 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_3 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x18x18x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1128xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_4 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_4 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_5 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_5 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<1x1x64x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1004xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_6 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret1, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret2, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret3, set=0, binding=4, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret4, set=0, binding=5, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret5, set=0, binding=6, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret6, set=0, binding=7, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_6 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> (tensor<1x8x8x64xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>)} | |
| hal.executable.binary attributes {data = dense<"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vector<2820xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_7 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_7 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_8 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_8 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<2512xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_9 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_9 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_10 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_10 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4216xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_11 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_11 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_12 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_12 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<2512xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_13 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_13 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_14 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_14 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<2512xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_15 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_15 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_16 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_16 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4076xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_17 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_17 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x1x256x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_18 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_18 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<2512xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_19 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_19 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<3x3x64x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_20 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_20 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>) -> tensor<1x8x8x64xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<2512xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_21 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_21 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x64xf32>, tensor<1x1x64x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000E003000020000000424D4F44180020001C001800140010000C000000000000000800040018000000B00200005C0000005000000044000000280000001400000004000000060000006D6F64756C6500000300000074020000500200003002000005000000D40100009001000044010000FC00000094000000010000005400000001000000140000000100000000000000050100000800040054FEFFFF0800000014000000E2FDFFFF04000000010000000200000016000000707265646963745F65785F64697370617463685F323100008CFEFFFF08000000140000001AFEFFFF04000000010000000200000016000000707265646963745F65785F64697370617463685F32310000C4FEFFFF080000004000000052FEFFFF040000000C00000001000000000000000100000000000000010000000000000000000000000000000000000000000000000000000000000014000000766D6C612E636F6E762E6633326633322E6633320000000028FFFFFF0800000024000000B6FEFFFF0400000005000000010000000000000001000000000000000000000010000000766D6C612E6275666665722E636F7079000000006CFFFFFF080000002800000078FFFFFF080000000C00000001000000010000000300000001000000000000000000000010000000766D6C612E6275666665722E7669657700000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000001000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000010000000300000002000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000CAFFFFFF040000000F00000021766D6C612E696E7465726661636500E6FFFFFF040000000C00000021766D6C612E627566666572000006000800040006000000040000000300000069333200050100000800000900000100010009004000000200090100000003000908000000040009400000000500090001000006000902000000070052000000800300008000000000010001805202000080030001C0000002000100018052000000800300008000000300010002805202000080030002C000000100010002805201000080010001000100038053040000800C00FFFF0400FFFF0400FFFF04000200040002000200FFFFFFFF1B0001C0030004000400050002C003000300050006000380030004000400060003000300000000000000000003000300030003000300030000005200000080030000C000000700010000805203000080050003C0000000C0000001000000540000000000"> : vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_22 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_22 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>, tensor<1x8x8x256xf32>) -> tensor<1x8x8x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4076xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_23 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_23 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x1x256x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1032xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_24 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_24 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x8x8x256xf32>, tensor<1x1x256x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1032xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_25 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret1, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret2, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_25 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> (tensor<1x4x4x128xf32>, tensor<1x1x1x2048xf32>, tensor<1x1x1x2048xf32>)} | |
| hal.executable.binary attributes {data = dense<"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vector<3160xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_26 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_26 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_27 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_27 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_28 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_28 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_29 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret1, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret2, set=0, binding=4, type="StorageBuffer", access="Write|Discard" | |
| hal.interface.binding @ret3, set=0, binding=5, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_29 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>)} | |
| hal.executable.binary attributes {data = dense<"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vector<6428xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_30 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_30 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_31 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_31 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000B40B000020000000424D4F441800240020001C001800140010000C00000000000800040018000000E8040000980000008000000074000000680000002C0000001400000004000000060000006D6F64756C65000004000000700400004804000028040000080400000D000000AC0300006803000024030000D8020000900200004C0200000C020000D001000094010000580100001C010000E4000000A80000000100000068000000010000002800000004000000540400003C040000240400001804000001000000000000008B02000006000800B0FCFFFF08000000140000001EFCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F33310000E8FCFFFF080000001400000056FCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F3331000020FDFFFF080000001C0000008EFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6D61782E6633320000000058FDFFFF0800000018000000C6FCFFFF040000000200000002000000020000000D000000766D6C612E737172742E6633320000008CFDFFFF080000001C000000FAFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6469762E66333200000000C4FDFFFF080000001C00000032FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6D756C2E66333200000000FCFDFFFF080000001C0000006AFDFFFF04000000030000000200000002000000020000000C000000766D6C612E7375622E6633320000000034FEFFFF080000001C000000A2FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6164642E663332000000006CFEFFFF0800000020000000DAFDFFFF0400000004000000020000000000000002000000000000000D000000766D6C612E74696C652E783332000000A8FEFFFF080000002000000016FEFFFF04000000040000000200000000000000020000000000000012000000766D6C612E62726F6164636173742E7833320000E8FEFFFF080000002400000056FEFFFF0400000005000000020000000000000002000000000000000000000010000000766D6C612E6275666665722E636F7079000000002CFFFFFF080000002800000038FFFFFF080000000C00000001000000020000000300000002000000000000000000000010000000766D6C612E6275666665722E766965770000000074FFFFFF080000002000000080FFFFFF080000000C0000000100000002000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000002000000010000000100000011000000766D6C612E6275666665722E636F6E7374000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000020000000300000003000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000AAFFFFFF040000000F00000021766D6C612E696E7465726661636500C6FFFFFF040000000C00000021766D6C612E62756666657200000000E2FFFFFF040000001100000021697265652E627974655F6275666665720006000800040006000000040000000300000069333200DAFFFFFFC4020000F6FFFFFFC402000000000A000A000000000004000A000000B404000000000A0008000000000004000A000000A40600008B02000008000009002000000100090002000002000901000000030009800000000400090400000005000B0000000001805201000080010001C0010001800B0100000002805201000080010002C0010002800B0200000003805201000080010003C0010003800B0300000004805201000080010004C00100048052000000800300008000000000010005805203000080030005C000000100010005805202000080010001000100068053060000800400FFFF0400FFFF04000A0003C00300030003000400068003000500050004000000520200008001000100010003805207000080030005C0068003800000520200008001000100010005805208000080030003C00680058000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002800300030003000400038003000500050004000000520200008001000100010007805209000080030005C003C0078000005202000080010002000100038053050000800400FFFF0000FFFF0100030004C0038004000000520200008001000200010004805207000080030003C002C00480000052020000800100020001000280520B000080020004C0028000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002C0030003000300040003800300050005000400000052020000800100010001000280520A000080030007C003C002800000520200008001000100010003805207000080030002C006C0038000005202000080010001000100028053050000800400FFFF0000FFFF0400060001C002800300050005000400000052020000800100010001000180520C000080030003C002C0018000005200000080030000C000000300010000805204000080050001C0000000C000000100000054000000040000009FF02737000200000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000200000000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0400000000000000"> : vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_32 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_32 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_33 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_33 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_34 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_34 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_35 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_35 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000D017000020000000424D4F441800240020001C001800140010000C00000000000800040018000000DC040000940000008000000074000000680000002C0000001400000004000000060000006D6F64756C650000040000006C0400004404000024040000040400000D000000A80300006403000020030000D40200008C0200004802000008020000CC010000900100005401000018010000E0000000A400000001000000640000000100000024000000030000004804000030040000180400000100000000000000B902000008000900B0FCFFFF08000000140000001EFCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F33350000E8FCFFFF080000001400000056FCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F3335000020FDFFFF080000001C0000008EFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6D61782E6633320000000058FDFFFF0800000018000000C6FCFFFF040000000200000002000000020000000D000000766D6C612E737172742E6633320000008CFDFFFF080000001C000000FAFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6469762E66333200000000C4FDFFFF080000001C00000032FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6D756C2E66333200000000FCFDFFFF080000001C0000006AFDFFFF04000000030000000200000002000000020000000C000000766D6C612E7375622E6633320000000034FEFFFF080000001C000000A2FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6164642E663332000000006CFEFFFF0800000020000000DAFDFFFF0400000004000000020000000000000002000000000000000D000000766D6C612E74696C652E783332000000A8FEFFFF080000002000000016FEFFFF04000000040000000200000000000000020000000000000012000000766D6C612E62726F6164636173742E7833320000E8FEFFFF080000002400000056FEFFFF0400000005000000020000000000000002000000000000000000000010000000766D6C612E6275666665722E636F7079000000002CFFFFFF080000002800000038FFFFFF080000000C00000001000000020000000300000002000000000000000000000010000000766D6C612E6275666665722E766965770000000074FFFFFF080000002000000080FFFFFF080000000C0000000100000002000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000002000000010000000100000011000000766D6C612E6275666665722E636F6E7374000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000020000000300000003000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000AAFFFFFF040000000F00000021766D6C612E696E7465726661636500C6FFFFFF040000000C00000021766D6C612E62756666657200000000E2FFFFFF040000001100000021697265652E627974655F6275666665720006000800040006000000040000000300000069333200F6FFFFFFEC02000000000A000A000000000004000A000000E002000000000A0008000000000004000A000000D00A0000B9020000080000090080000001000900080000020009010000000300090200000004000900020000050009040000000600090300000007000B0000000001805201000080010001C0010001800B0100000002805201000080010002C0010002800B0200000003805201000080010003C00100038052000000800300008000000000010004805203000080030004C0000001000100048052000000800300008000000300010005805203000080030005C0000001000100058052000000800300008000000400010006805203000080030006C000000100010006805202000080010001000100078053060000800400FFFF0400FFFF04000A0002C00300030003000500078003000600060005000000520200008001000100010002805207000080030004C0078002800000520200008001000100010004805208000080030002C00780048000005202000080010001000100028053060000800400FFFF0400FFFF04000A0001800300030003000500028003000600060005000000520200008001000100010008805209000080030004C002C0088000005202000080010002000100028053050000800400FFFF0000FFFF0100030003C0028005000000520200008001000200010003805207000080030002C001C00380000052020000800100020001000180520B000080020003C0018000005202000080010001000100028053060000800400FFFF0400FFFF04000A0001C0030003000300050002800300060006000500000052020000800100010001000180520A000080030008C002C001800000520200008001000100010002805207000080030001C007C002800000520200008001000100010001805207000080030005C002C00180000052020000800100010001000280520C000080030001C006C0028000005200000080030000C000000700010000805204000080050002C0000000C0000001000000540000000000040000009FF02737000800000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000800000000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F"> : vector<6124xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_36 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_36 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_37 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_37 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_38 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_38 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_39 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_39 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_40 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_40 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_41 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_41 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<6124xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_42 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_42 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x1x512x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_43 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_43 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_44 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_44 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<3x3x128x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_45 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_45 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>) -> tensor<1x4x4x128xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<3024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_46 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_46 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x128xf32>, tensor<1x1x128x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_47 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_47 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>, tensor<1x4x4x512xf32>) -> tensor<1x4x4x512xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<6124xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_48 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_48 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x1x512x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_49 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_49 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x4x4x512xf32>, tensor<1x1x512x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1024xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_50 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_50 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_51 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_51 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1012xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_52 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_52 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000B40F000020000000424D4F441800240020001C001800140010000C00000000000800040018000000E8040000980000008000000074000000680000002C0000001400000004000000060000006D6F64756C65000004000000700400004804000028040000080400000D000000AC0300006803000024030000D8020000900200004C0200000C020000D001000094010000580100001C010000E4000000A80000000100000068000000010000002800000004000000540400003C040000240400001804000001000000000000008B02000006000800B0FCFFFF08000000140000001EFCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F35320000E8FCFFFF080000001400000056FCFFFF04000000010000000300000016000000707265646963745F65785F64697370617463685F3532000020FDFFFF080000001C0000008EFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6D61782E6633320000000058FDFFFF0800000018000000C6FCFFFF040000000200000002000000020000000D000000766D6C612E737172742E6633320000008CFDFFFF080000001C000000FAFCFFFF04000000030000000200000002000000020000000C000000766D6C612E6469762E66333200000000C4FDFFFF080000001C00000032FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6D756C2E66333200000000FCFDFFFF080000001C0000006AFDFFFF04000000030000000200000002000000020000000C000000766D6C612E7375622E6633320000000034FEFFFF080000001C000000A2FDFFFF04000000030000000200000002000000020000000C000000766D6C612E6164642E663332000000006CFEFFFF0800000020000000DAFDFFFF0400000004000000020000000000000002000000000000000D000000766D6C612E74696C652E783332000000A8FEFFFF080000002000000016FEFFFF04000000040000000200000000000000020000000000000012000000766D6C612E62726F6164636173742E7833320000E8FEFFFF080000002400000056FEFFFF0400000005000000020000000000000002000000000000000000000010000000766D6C612E6275666665722E636F7079000000002CFFFFFF080000002800000038FFFFFF080000000C00000001000000020000000300000002000000000000000000000010000000766D6C612E6275666665722E766965770000000074FFFFFF080000002000000080FFFFFF080000000C0000000100000002000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000002000000010000000100000011000000766D6C612E6275666665722E636F6E7374000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000020000000300000003000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000AAFFFFFF040000000F00000021766D6C612E696E7465726661636500C6FFFFFF040000000C00000021766D6C612E62756666657200000000E2FFFFFF040000001100000021697265652E627974655F6275666665720006000800040006000000040000000300000069333200DAFFFFFFC4020000F6FFFFFFC402000000000A000A000000000004000A000000B406000000000A0008000000000004000A000000A40A00008B02000008000009001000000100090004000002000901000000030009000100000400090200000005000B0000000001805201000080010001C0010001800B0100000002805201000080010002C0010002800B0200000003805201000080010003C0010003800B0300000004805201000080010004C00100048052000000800300008000000000010005805203000080030005C000000100010005805202000080010001000100068053060000800400FFFF0400FFFF04000A0003C00300030003000400068003000500050004000000520200008001000100010003805207000080030005C0068003800000520200008001000100010005805208000080030003C00680058000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002800300030003000400038003000500050004000000520200008001000100010007805209000080030005C003C0078000005202000080010002000100038053050000800400FFFF0000FFFF0100030004C0038004000000520200008001000200010004805207000080030003C002C00480000052020000800100020001000280520B000080020004C0028000005202000080010001000100038053060000800400FFFF0400FFFF04000A0002C0030003000300040003800300050005000400000052020000800100010001000280520A000080030007C003C002800000520200008001000100010003805207000080030002C006C0038000005202000080010001000100028053050000800400FFFF0000FFFF0400060001C002800300050005000400000052020000800100010001000180520C000080030003C002C0018000005200000080030000C000000300010000805204000080050001C0000000C000000100000054000000040000009FF027370004000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000400000000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0000803F0400000000000000"> : vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_53 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_53 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_54 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_54 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<10352xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_55 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_55 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_56 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_56 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_57 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_57 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1012xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_58 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_58 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_59 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_59 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_60 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_60 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<10212xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_61 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_61 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_62 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_62 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_63 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_63 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1012xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_64 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_64 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_65 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_65 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<1x1x256x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_66 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg2, set=0, binding=2, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=3, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_66 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>, tensor<1x2x2x1024xf32>) -> tensor<1x2x2x1024xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<10212xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_67 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_67 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x1024xf32>, tensor<1x1x1024x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"0x10000000564D4C4100000600080004000600000004000000E003000020000000424D4F44180020001C001800140010000C000000000000000800040018000000B00200005C0000005000000044000000280000001400000004000000060000006D6F64756C6500000300000074020000500200003002000005000000D40100009001000044010000FC00000094000000010000005400000001000000140000000100000000000000050100000800040054FEFFFF0800000014000000E2FDFFFF04000000010000000200000016000000707265646963745F65785F64697370617463685F363700008CFEFFFF08000000140000001AFEFFFF04000000010000000200000016000000707265646963745F65785F64697370617463685F36370000C4FEFFFF080000004000000052FEFFFF040000000C00000001000000000000000100000000000000010000000000000000000000000000000000000000000000000000000000000014000000766D6C612E636F6E762E6633326633322E6633320000000028FFFFFF0800000024000000B6FEFFFF0400000005000000010000000000000001000000000000000000000010000000766D6C612E6275666665722E636F7079000000006CFFFFFF080000002800000078FFFFFF080000000C00000001000000010000000300000001000000000000000000000010000000766D6C612E6275666665722E7669657700000000B4FFFFFF0800000020000000C0FFFFFF080000000C0000000100000001000000010000000000000011000000766D6C612E6275666665722E616C6C6F63000000F4FFFFFF100000003000000008000C000800040008000000080000000C00000001000000010000000300000002000000000000000000000016000000766D6C612E696E746572666163652E62696E64696E670000CAFFFFFF040000000F00000021766D6C612E696E7465726661636500E6FFFFFF040000000C00000021766D6C612E627566666572000006000800040006000000040000000300000069333200050100000800000900001000010009001000000200090040000003000901000000040009020000000500090004000006000900010000070052000000800300008000000000010001805202000080030001C0000003000100018052000000800300008000000400010002805202000080030002C000000100010002805201000080010002000100038053040000800C00FFFF0400FFFF0400FFFF04000200040002000200FFFFFFFF1B0001C0040005000500060002C004000400060007000380040005000500070004000400000000000000000004000400040004000400040000005200000080030000C000000500010000805203000080050003C0000000C0000002000000540000000000"> : vector<1020xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_68 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_68 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<4048xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_69 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @arg1, set=0, binding=1, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=2, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_69 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>, tensor<3x3x256x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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vector<1012xi8>, format = 1447906369 : i32} { | |
| } | |
| } | |
| hal.executable @predict_ex_dispatch_70 attributes {sym_visibility = "private"} { | |
| hal.interface @legacy_io { | |
| hal.interface.binding @arg0, set=0, binding=0, type="StorageBuffer", access="Read" | |
| hal.interface.binding @ret0, set=0, binding=1, type="StorageBuffer", access="Write|Discard" | |
| } | |
| hal.executable.entry_point @predict_ex_dispatch_70 attributes {interface = @legacy_io, ordinal = 0 : i32, signature = (tensor<1x2x2x256xf32>) -> tensor<1x2x2x256xf32>} | |
| hal.executable.binary attributes {data = dense<"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|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment