Last active
July 4, 2021 19:14
-
-
Save iviarcio/a53cf5feef166885d998b6dee93b64ef to your computer and use it in GitHub Desktop.
resnet50 low-level intemediate representation compiled with int16 quantization
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
function resnet50 | |
declare { | |
%gpu_0_pred_b = WeightVar i16[S:0.000001907 O:0][-0.062,0.062]<1000> const // size: 2000 // Users: @in 205 | |
%gpu_0_pred_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<2048 x 1000> const // size: 4096000 // Users: @in 205 | |
%gpu_0_conv1_bias_constfold = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<64> const // size: 128 // Users: @in 3 | |
%gpu_0_res2_0_branch2c_bias_constfold = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 154 | |
%gpu_0_res3_0_branch2a_bias_constfold = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 85 | |
%gpu_0_res3_0_branch2c_bias_constfold = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<512> const // size: 1024 // Users: @in 192 | |
%gpu_0_res4_0_branch2c_bias_constfold = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<1024> const // size: 2048 // Users: @in 157 | |
%gpu_0_res5_0_branch2c_bias_constfold = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<2048> const // size: 4096 // Users: @in 195 | |
%gpu_0_res2_0_branch2c_bias_constfold__1 = WeightVar i16[S:0.000488281 O:0][-16.000,16.000]<256> const // size: 512 // Users: @in 17 | |
%gpu_0_conv1_bias_constfold__1 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<64> const // size: 128 // Users: @in 9 | |
%gpu_0_conv1_bias_constfold__2 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<64> const // size: 128 // Users: @in 11 | |
%gpu_0_res2_0_branch2c_bias_constfold__2 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<256> const // size: 512 // Users: @in 14 | |
%gpu_0_conv1_bias_constfold__3 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<64> const // size: 128 // Users: @in 25 | |
%gpu_0_conv1_bias_constfold__4 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<64> const // size: 128 // Users: @in 27 | |
%gpu_0_res2_0_branch2c_bias_constfold__3 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 30 | |
%gpu_0_conv1_bias_constfold__5 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<64> const // size: 128 // Users: @in 36 | |
%gpu_0_conv1_bias_constfold__6 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<64> const // size: 128 // Users: @in 38 | |
%gpu_0_res2_0_branch2c_bias_constfold__4 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 41 | |
%gpu_0_res3_0_branch2c_bias_constfold__1 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<512> const // size: 1024 // Users: @in 55 | |
%gpu_0_res3_0_branch2a_bias_constfold__1 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<128> const // size: 256 // Users: @in 47 | |
%gpu_0_res3_0_branch2a_bias_constfold__2 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 49 | |
%gpu_0_res3_0_branch2c_bias_constfold__2 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<512> const // size: 1024 // Users: @in 52 | |
%gpu_0_res3_0_branch2a_bias_constfold__3 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 61 | |
%gpu_0_res3_0_branch2a_bias_constfold__4 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<128> const // size: 256 // Users: @in 63 | |
%gpu_0_res3_0_branch2c_bias_constfold__3 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<512> const // size: 1024 // Users: @in 66 | |
%gpu_0_res3_0_branch2a_bias_constfold__5 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 72 | |
%gpu_0_res3_0_branch2a_bias_constfold__6 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 74 | |
%gpu_0_res3_0_branch2c_bias_constfold__4 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<512> const // size: 1024 // Users: @in 77 | |
%gpu_0_res3_0_branch2a_bias_constfold__7 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<128> const // size: 256 // Users: @in 83 | |
%gpu_0_res3_0_branch2c_bias_constfold__5 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512> const // size: 1024 // Users: @in 88 | |
%gpu_0_res4_0_branch2c_bias_constfold__1 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<1024> const // size: 2048 // Users: @in 102 | |
%gpu_0_res2_0_branch2c_bias_constfold__5 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 94 | |
%gpu_0_res2_0_branch2c_bias_constfold__6 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256> const // size: 512 // Users: @in 96 | |
%gpu_0_res4_0_branch2c_bias_constfold__2 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<1024> const // size: 2048 // Users: @in 99 | |
%gpu_0_res2_0_branch2c_bias_constfold__7 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<256> const // size: 512 // Users: @in 108 | |
%gpu_0_res2_0_branch2c_bias_constfold__8 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<256> const // size: 512 // Users: @in 110 | |
%gpu_0_res4_0_branch2c_bias_constfold__3 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<1024> const // size: 2048 // Users: @in 113 | |
%gpu_0_res2_0_branch2c_bias_constfold__9 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<256> const // size: 512 // Users: @in 119 | |
%gpu_0_res2_0_branch2c_bias_constfold__10 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 121 | |
%gpu_0_res4_0_branch2c_bias_constfold__4 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<1024> const // size: 2048 // Users: @in 124 | |
%gpu_0_res2_0_branch2c_bias_constfold__11 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256> const // size: 512 // Users: @in 130 | |
%gpu_0_res2_0_branch2c_bias_constfold__12 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 132 | |
%gpu_0_res4_0_branch2c_bias_constfold__5 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<1024> const // size: 2048 // Users: @in 135 | |
%gpu_0_res2_0_branch2c_bias_constfold__13 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 141 | |
%gpu_0_res2_0_branch2c_bias_constfold__14 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<256> const // size: 512 // Users: @in 143 | |
%gpu_0_res4_0_branch2c_bias_constfold__6 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<1024> const // size: 2048 // Users: @in 146 | |
%gpu_0_res2_0_branch2c_bias_constfold__15 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256> const // size: 512 // Users: @in 152 | |
%gpu_0_res5_0_branch2c_bias_constfold__1 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<2048> const // size: 4096 // Users: @in 171 | |
%gpu_0_res3_0_branch2c_bias_constfold__6 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512> const // size: 1024 // Users: @in 163 | |
%gpu_0_res3_0_branch2c_bias_constfold__7 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<512> const // size: 1024 // Users: @in 165 | |
%gpu_0_res5_0_branch2c_bias_constfold__2 = WeightVar i16[S:0.000488281 O:0][-16.000,16.000]<2048> const // size: 4096 // Users: @in 168 | |
%gpu_0_res3_0_branch2c_bias_constfold__8 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512> const // size: 1024 // Users: @in 179 | |
%gpu_0_res3_0_branch2c_bias_constfold__9 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<512> const // size: 1024 // Users: @in 181 | |
%gpu_0_res5_0_branch2c_bias_constfold__3 = WeightVar i16[S:0.000488281 O:0][-16.000,16.000]<2048> const // size: 4096 // Users: @in 184 | |
%gpu_0_res3_0_branch2c_bias_constfold__10 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<512> const // size: 1024 // Users: @in 190 | |
%gpu_0_res2_0_branch1_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256 x 64 x 1 x 1> const // size: 32768 // Users: @in 17 | |
%gpu_0_res2_0_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<64 x 64 x 1 x 1> const // size: 8192 // Users: @in 9 | |
%gpu_0_res2_0_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256 x 64 x 1 x 1> const // size: 32768 // Users: @in 14 | |
%gpu_0_res2_1_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<64 x 256 x 1 x 1> const // size: 32768 // Users: @in 25 | |
%gpu_0_res2_1_branch2c_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 64 x 1 x 1> const // size: 32768 // Users: @in 30 | |
%gpu_0_res2_2_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<64 x 256 x 1 x 1> const // size: 32768 // Users: @in 36 | |
%gpu_0_res2_2_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<256 x 64 x 1 x 1> const // size: 32768 // Users: @in 41 | |
%gpu_0_res3_0_branch1_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512 x 256 x 1 x 1> const // size: 262144 // Users: @in 55 | |
%gpu_0_res3_0_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<128 x 256 x 1 x 1> const // size: 65536 // Users: @in 47 | |
%gpu_0_res3_0_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512 x 128 x 1 x 1> const // size: 131072 // Users: @in 52 | |
%gpu_0_res3_1_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<128 x 512 x 1 x 1> const // size: 131072 // Users: @in 61 | |
%gpu_0_res3_1_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512 x 128 x 1 x 1> const // size: 131072 // Users: @in 66 | |
%gpu_0_res3_2_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<128 x 512 x 1 x 1> const // size: 131072 // Users: @in 72 | |
%gpu_0_res3_2_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512 x 128 x 1 x 1> const // size: 131072 // Users: @in 77 | |
%gpu_0_res3_3_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<128 x 512 x 1 x 1> const // size: 131072 // Users: @in 83 | |
%gpu_0_res3_3_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<512 x 128 x 1 x 1> const // size: 131072 // Users: @in 88 | |
%gpu_0_res4_0_branch1_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<1024 x 512 x 1 x 1> const // size: 1048576 // Users: @in 102 | |
%gpu_0_res4_0_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 512 x 1 x 1> const // size: 262144 // Users: @in 94 | |
%gpu_0_res4_0_branch2c_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 99 | |
%gpu_0_res4_1_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 1024 x 1 x 1> const // size: 524288 // Users: @in 108 | |
%gpu_0_res4_1_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 113 | |
%gpu_0_res4_2_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 1024 x 1 x 1> const // size: 524288 // Users: @in 119 | |
%gpu_0_res4_2_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 124 | |
%gpu_0_res4_3_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 1024 x 1 x 1> const // size: 524288 // Users: @in 130 | |
%gpu_0_res4_3_branch2c_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 135 | |
%gpu_0_res4_4_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 1024 x 1 x 1> const // size: 524288 // Users: @in 141 | |
%gpu_0_res4_4_branch2c_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 146 | |
%gpu_0_res4_5_branch2a_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 1024 x 1 x 1> const // size: 524288 // Users: @in 152 | |
%gpu_0_res4_5_branch2c_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<1024 x 256 x 1 x 1> const // size: 524288 // Users: @in 157 | |
%gpu_0_res5_0_branch1_w__2 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<2048 x 1024 x 1 x 1> const // size: 4194304 // Users: @in 171 | |
%gpu_0_res5_0_branch2a_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<512 x 1024 x 1 x 1> const // size: 1048576 // Users: @in 163 | |
%gpu_0_res5_0_branch2c_w__2 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<2048 x 512 x 1 x 1> const // size: 2097152 // Users: @in 168 | |
%gpu_0_res5_1_branch2a_w__2 = WeightVar i16[S:0.000003815 O:0][-0.125,0.125]<512 x 2048 x 1 x 1> const // size: 2097152 // Users: @in 179 | |
%gpu_0_res5_1_branch2c_w__2 = WeightVar i16[S:0.000122070 O:0][-4.000,4.000]<2048 x 512 x 1 x 1> const // size: 2097152 // Users: @in 184 | |
%gpu_0_res5_2_branch2a_w__2 = WeightVar i16[S:0.000007629 O:0][-0.250,0.250]<512 x 2048 x 1 x 1> const // size: 2097152 // Users: @in 190 | |
%gpu_0_res5_2_branch2c_w__2 = WeightVar i16[S:0.000244141 O:0][-8.000,8.000]<2048 x 512 x 1 x 1> const // size: 2097152 // Users: @in 195 | |
%gpu_0_conv1_w__2 = WeightVar i16[S:0.000061035 O:0][-2.000,2.000]<64 x 3 x 7 x 7> const // size: 18816 // Users: @in 3 | |
%gpu_0_res2_0_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<64 x 64 x 3 x 3> const // size: 73728 // Users: @in 11 | |
%gpu_0_res2_1_branch2b_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<64 x 64 x 3 x 3> const // size: 73728 // Users: @in 27 | |
%gpu_0_res2_2_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<64 x 64 x 3 x 3> const // size: 73728 // Users: @in 38 | |
%gpu_0_res3_0_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<128 x 128 x 3 x 3> const // size: 294912 // Users: @in 49 | |
%gpu_0_res3_1_branch2b_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<128 x 128 x 3 x 3> const // size: 294912 // Users: @in 63 | |
%gpu_0_res3_2_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<128 x 128 x 3 x 3> const // size: 294912 // Users: @in 74 | |
%gpu_0_res3_3_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<128 x 128 x 3 x 3> const // size: 294912 // Users: @in 85 | |
%gpu_0_res4_0_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 96 | |
%gpu_0_res4_1_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 110 | |
%gpu_0_res4_2_branch2b_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 121 | |
%gpu_0_res4_3_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 132 | |
%gpu_0_res4_4_branch2b_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 143 | |
%gpu_0_res4_5_branch2b_w__2 = WeightVar i16[S:0.000030518 O:0][-1.000,1.000]<256 x 256 x 3 x 3> const // size: 1179648 // Users: @in 154 | |
%gpu_0_res5_0_branch2b_w__2 = WeightVar i16[S:0.000015259 O:0][-0.500,0.500]<512 x 512 x 3 x 3> const // size: 4718592 // Users: @in 165 | |
%gpu_0_res5_1_branch2b_w__2 = WeightVar i16[S:0.000007629 O:0][-0.250,0.250]<512 x 512 x 3 x 3> const // size: 4718592 // Users: @in 181 | |
%gpu_0_res5_2_branch2b_w__2 = WeightVar i16[S:0.000007629 O:0][-0.250,0.250]<512 x 512 x 3 x 3> const // size: 4718592 // Users: @in 192 | |
%gpu_0_data = WeightVar float<1 x 3 x 224 x 224> mutable // size: 602112 // Users: @in 1 | |
%gpu_0_softmax = WeightVar float<1 x 1000> mutable // size: 4000 // Users: @out 210 | |
; size = 51667056 bytes | |
} | |
code { | |
0 %gpu_0_conv1_quantize_res = allocactivation { Ty: i16[S:0.000061035 O:0][-2.000,2.000]<1 x 3 x 224 x 224>} // size: 301056 // Users: @out 4, @in 3, @out 1 | |
1 %gpu_0_conv1_quantize = quantize @out %gpu_0_conv1_quantize_res, @in %gpu_0_data | |
2 %gpu_0_conv1__10_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 64 x 112 x 112>} // size: 1605632 // Users: @in 6, @out 7, @out 3 | |
3 %gpu_0_conv1__10 = convolution @out %gpu_0_conv1__10_res, @in %gpu_0_conv1_quantize_res, @in %gpu_0_conv1_w__2, @in %gpu_0_conv1_bias_constfold { Kernels: [7, 7], Strides: [2, 2], Pads: [3, 3, 3, 3], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
4 %dealloc_gpu_0_conv1_quantize_res = deallocactivation @out %gpu_0_conv1_quantize_res // size: 301056 | |
5 %gpu_0_pool1__4_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 6, @out 18, @in 17, @in 9 | |
6 %gpu_0_pool1__1 = maxpool @out %gpu_0_pool1__4_res, @in %gpu_0_conv1__10_res { Kernels: [3, 3], Strides: [2, 2], Pads: [1, 1, 1, 1], Layout: 1} | |
7 %dealloc_gpu_0_conv1__10_res = deallocactivation @out %gpu_0_conv1__10_res // size: 1605632 | |
8 %gpu_0_res2_0_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 12, @in 11, @out 9 | |
9 %gpu_0_res2_0_branch2a__5 = convolution @out %gpu_0_res2_0_branch2a__5_res, @in %gpu_0_pool1__4_res, @in %gpu_0_res2_0_branch2a_w__2, @in %gpu_0_conv1_bias_constfold__1 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
10 %gpu_0_res2_0_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 15, @in 14, @out 11 | |
11 %gpu_0_res2_0_branch2b__5 = convolution @out %gpu_0_res2_0_branch2b__5_res, @in %gpu_0_res2_0_branch2a__5_res, @in %gpu_0_res2_0_branch2b_w__2, @in %gpu_0_conv1_bias_constfold__2 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
12 %dealloc_gpu_0_res2_0_branch2a__5_res = deallocactivation @out %gpu_0_res2_0_branch2a__5_res // size: 401408 | |
13 %gpu_0_res2_0_branch2c__6_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 56 x 56>} // size: 1605632 // Users: @out 20, @in 19, @out 14 | |
14 %gpu_0_res2_0_branch2c__6 = convolution @out %gpu_0_res2_0_branch2c__6_res, @in %gpu_0_res2_0_branch2b__5_res, @in %gpu_0_res2_0_branch2c_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__2 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
15 %dealloc_gpu_0_res2_0_branch2b__5_res = deallocactivation @out %gpu_0_res2_0_branch2b__5_res // size: 401408 | |
16 %gpu_0_res2_0_branch1__4_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 56 x 56>} // size: 1605632 // Users: @in 22, @out 19, @out 23, @in 19, @out 17 | |
17 %gpu_0_res2_0_branch1__4 = convolution @out %gpu_0_res2_0_branch1__4_res, @in %gpu_0_pool1__4_res, @in %gpu_0_res2_0_branch1_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__1 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
18 %dealloc_gpu_0_pool1__4_res = deallocactivation @out %gpu_0_pool1__4_res // size: 401408 | |
19 %gpu_0_res2_0_branch2c_bn__6 = elementadd @out %gpu_0_res2_0_branch1__4_res, @in %gpu_0_res2_0_branch2c__6_res, @in %gpu_0_res2_0_branch1__4_res | |
20 %dealloc_gpu_0_res2_0_branch2c__6_res = deallocactivation @out %gpu_0_res2_0_branch2c__6_res // size: 1605632 | |
21 %gpu_0_res2_0_branch2c_bn__7_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 56 x 56>} // size: 1605632 // Users: @out 33, @in 32, @in 25, @out 22 | |
22 %gpu_0_res2_0_branch2c_bn__7 = relu @out %gpu_0_res2_0_branch2c_bn__7_res, @in %gpu_0_res2_0_branch1__4_res | |
23 %dealloc_gpu_0_res2_0_branch1__4_res = deallocactivation @out %gpu_0_res2_0_branch1__4_res // size: 1605632 | |
24 %gpu_0_res2_1_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 28, @in 27, @out 25 | |
25 %gpu_0_res2_1_branch2a__5 = convolution @out %gpu_0_res2_1_branch2a__5_res, @in %gpu_0_res2_0_branch2c_bn__7_res, @in %gpu_0_res2_1_branch2a_w__2, @in %gpu_0_conv1_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
26 %gpu_0_res2_1_branch2b__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 31, @in 30, @out 27 | |
27 %gpu_0_res2_1_branch2b__5 = convolution @out %gpu_0_res2_1_branch2b__5_res, @in %gpu_0_res2_1_branch2a__5_res, @in %gpu_0_res2_1_branch2b_w__2, @in %gpu_0_conv1_bias_constfold__4 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
28 %dealloc_gpu_0_res2_1_branch2a__5_res = deallocactivation @out %gpu_0_res2_1_branch2a__5_res // size: 401408 | |
29 %gpu_0_res2_1_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 56 x 56>} // size: 1605632 // Users: @in 32, @out 30, @in 43, @in 36, @out 34, @out 44, @in 34, @out 32 | |
30 %gpu_0_res2_1_branch2c__6 = convolution @out %gpu_0_res2_1_branch2c_bn__6_res, @in %gpu_0_res2_1_branch2b__5_res, @in %gpu_0_res2_1_branch2c_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
31 %dealloc_gpu_0_res2_1_branch2b__5_res = deallocactivation @out %gpu_0_res2_1_branch2b__5_res // size: 401408 | |
32 %gpu_0_res2_1_branch2c_bn__6 = elementadd @out %gpu_0_res2_1_branch2c_bn__6_res, @in %gpu_0_res2_1_branch2c_bn__6_res, @in %gpu_0_res2_0_branch2c_bn__7_res | |
33 %dealloc_gpu_0_res2_0_branch2c_bn__7_res = deallocactivation @out %gpu_0_res2_0_branch2c_bn__7_res // size: 1605632 | |
34 %gpu_0_res2_1_branch2c_bn__7 = relu @out %gpu_0_res2_1_branch2c_bn__6_res, @in %gpu_0_res2_1_branch2c_bn__6_res | |
35 %gpu_0_res2_2_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 39, @in 38, @out 36 | |
36 %gpu_0_res2_2_branch2a__5 = convolution @out %gpu_0_res2_2_branch2a__5_res, @in %gpu_0_res2_1_branch2c_bn__6_res, @in %gpu_0_res2_2_branch2a_w__2, @in %gpu_0_conv1_bias_constfold__5 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
37 %gpu_0_res2_2_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 64 x 56 x 56>} // size: 401408 // Users: @out 42, @in 41, @out 38 | |
38 %gpu_0_res2_2_branch2b__5 = convolution @out %gpu_0_res2_2_branch2b__5_res, @in %gpu_0_res2_2_branch2a__5_res, @in %gpu_0_res2_2_branch2b_w__2, @in %gpu_0_conv1_bias_constfold__6 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
39 %dealloc_gpu_0_res2_2_branch2a__5_res = deallocactivation @out %gpu_0_res2_2_branch2a__5_res // size: 401408 | |
40 %gpu_0_res2_2_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 56 x 56>} // size: 1605632 // Users: @in 43, @out 41, @in 55, @in 47, @out 45, @out 56, @in 45, @out 43 | |
41 %gpu_0_res2_2_branch2c__6 = convolution @out %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res2_2_branch2b__5_res, @in %gpu_0_res2_2_branch2c_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__4 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
42 %dealloc_gpu_0_res2_2_branch2b__5_res = deallocactivation @out %gpu_0_res2_2_branch2b__5_res // size: 401408 | |
43 %gpu_0_res2_2_branch2c_bn__6 = elementadd @out %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res2_1_branch2c_bn__6_res | |
44 %dealloc_gpu_0_res2_1_branch2c_bn__6_res = deallocactivation @out %gpu_0_res2_1_branch2c_bn__6_res // size: 1605632 | |
45 %gpu_0_res2_2_branch2c_bn__7 = relu @out %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res2_2_branch2c_bn__6_res | |
46 %gpu_0_res3_0_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 128 x 56 x 56>} // size: 802816 // Users: @out 50, @in 49, @out 47 | |
47 %gpu_0_res3_0_branch2a__5 = convolution @out %gpu_0_res3_0_branch2a__5_res, @in %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res3_0_branch2a_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__1 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
48 %gpu_0_res3_0_branch2b__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 53, @in 52, @out 49 | |
49 %gpu_0_res3_0_branch2b__5 = convolution @out %gpu_0_res3_0_branch2b__5_res, @in %gpu_0_res3_0_branch2a__5_res, @in %gpu_0_res3_0_branch2b_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__2 { Kernels: [3, 3], Strides: [2, 2], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
50 %dealloc_gpu_0_res3_0_branch2a__5_res = deallocactivation @out %gpu_0_res3_0_branch2a__5_res // size: 802816 | |
51 %gpu_0_res3_2_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000976562 O:0][-32.000,31.999]<1 x 512 x 28 x 28>} // size: 802816 // Users: @in 57, @out 52, @in 59, @out 57, @in 68, @in 61, @out 59, @in 70, @out 68, @in 79, @in 72, @out 70, @in 90, @in 83, @out 81, @out 91, @in 81, @out 79 | |
52 %gpu_0_res3_0_branch2c__6 = convolution @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_0_branch2b__5_res, @in %gpu_0_res3_0_branch2c_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__2 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
53 %dealloc_gpu_0_res3_0_branch2b__5_res = deallocactivation @out %gpu_0_res3_0_branch2b__5_res // size: 200704 | |
54 %gpu_0_res3_0_branch1__4_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 512 x 28 x 28>} // size: 802816 // Users: @out 58, @in 57, @out 55 | |
55 %gpu_0_res3_0_branch1__4 = convolution @out %gpu_0_res3_0_branch1__4_res, @in %gpu_0_res2_2_branch2c_bn__6_res, @in %gpu_0_res3_0_branch1_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__1 { Kernels: [1, 1], Strides: [2, 2], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
56 %dealloc_gpu_0_res2_2_branch2c_bn__6_res = deallocactivation @out %gpu_0_res2_2_branch2c_bn__6_res // size: 1605632 | |
57 %gpu_0_res3_0_branch2c_bn__6 = elementadd @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_0_branch1__4_res | |
58 %dealloc_gpu_0_res3_0_branch1__4_res = deallocactivation @out %gpu_0_res3_0_branch1__4_res // size: 802816 | |
59 %gpu_0_res3_0_branch2c_bn__7 = relu @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
60 %gpu_0_res3_1_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 64, @in 63, @out 61 | |
61 %gpu_0_res3_1_branch2a__5 = convolution @out %gpu_0_res3_1_branch2a__5_res, @in %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_1_branch2a_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
62 %gpu_0_res3_1_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 67, @in 66, @out 63 | |
63 %gpu_0_res3_1_branch2b__5 = convolution @out %gpu_0_res3_1_branch2b__5_res, @in %gpu_0_res3_1_branch2a__5_res, @in %gpu_0_res3_1_branch2b_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__4 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
64 %dealloc_gpu_0_res3_1_branch2a__5_res = deallocactivation @out %gpu_0_res3_1_branch2a__5_res // size: 200704 | |
65 %gpu_0_res3_1_branch2c__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 512 x 28 x 28>} // size: 802816 // Users: @out 69, @in 68, @out 66 | |
66 %gpu_0_res3_1_branch2c__6 = convolution @out %gpu_0_res3_1_branch2c__6_res, @in %gpu_0_res3_1_branch2b__5_res, @in %gpu_0_res3_1_branch2c_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
67 %dealloc_gpu_0_res3_1_branch2b__5_res = deallocactivation @out %gpu_0_res3_1_branch2b__5_res // size: 200704 | |
68 %gpu_0_res3_1_branch2c_bn__6 = elementadd @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_1_branch2c__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
69 %dealloc_gpu_0_res3_1_branch2c__6_res = deallocactivation @out %gpu_0_res3_1_branch2c__6_res // size: 802816 | |
70 %gpu_0_res3_1_branch2c_bn__7 = relu @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
71 %gpu_0_res3_2_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 75, @in 74, @out 72 | |
72 %gpu_0_res3_2_branch2a__5 = convolution @out %gpu_0_res3_2_branch2a__5_res, @in %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2a_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__5 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
73 %gpu_0_res3_2_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 78, @in 77, @out 74 | |
74 %gpu_0_res3_2_branch2b__5 = convolution @out %gpu_0_res3_2_branch2b__5_res, @in %gpu_0_res3_2_branch2a__5_res, @in %gpu_0_res3_2_branch2b_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__6 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
75 %dealloc_gpu_0_res3_2_branch2a__5_res = deallocactivation @out %gpu_0_res3_2_branch2a__5_res // size: 200704 | |
76 %gpu_0_res3_2_branch2c__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 512 x 28 x 28>} // size: 802816 // Users: @out 80, @in 79, @out 77 | |
77 %gpu_0_res3_2_branch2c__6 = convolution @out %gpu_0_res3_2_branch2c__6_res, @in %gpu_0_res3_2_branch2b__5_res, @in %gpu_0_res3_2_branch2c_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__4 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
78 %dealloc_gpu_0_res3_2_branch2b__5_res = deallocactivation @out %gpu_0_res3_2_branch2b__5_res // size: 200704 | |
79 %gpu_0_res3_2_branch2c_bn__6 = elementadd @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
80 %dealloc_gpu_0_res3_2_branch2c__6_res = deallocactivation @out %gpu_0_res3_2_branch2c__6_res // size: 802816 | |
81 %gpu_0_res3_2_branch2c_bn__7 = relu @out %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
82 %gpu_0_res3_3_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 86, @in 85, @out 83 | |
83 %gpu_0_res3_3_branch2a__5 = convolution @out %gpu_0_res3_3_branch2a__5_res, @in %gpu_0_res3_2_branch2c_bn__6_res, @in %gpu_0_res3_3_branch2a_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold__7 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
84 %gpu_0_res3_3_branch2b__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 128 x 28 x 28>} // size: 200704 // Users: @out 89, @in 88, @out 85 | |
85 %gpu_0_res3_3_branch2b__5 = convolution @out %gpu_0_res3_3_branch2b__5_res, @in %gpu_0_res3_3_branch2a__5_res, @in %gpu_0_res3_3_branch2b_w__2, @in %gpu_0_res3_0_branch2a_bias_constfold { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
86 %dealloc_gpu_0_res3_3_branch2a__5_res = deallocactivation @out %gpu_0_res3_3_branch2a__5_res // size: 200704 | |
87 %gpu_0_res3_3_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 512 x 28 x 28>} // size: 802816 // Users: @in 90, @out 88, @in 102, @in 94, @out 92, @out 103, @in 92, @out 90 | |
88 %gpu_0_res3_3_branch2c__6 = convolution @out %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res3_3_branch2b__5_res, @in %gpu_0_res3_3_branch2c_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__5 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
89 %dealloc_gpu_0_res3_3_branch2b__5_res = deallocactivation @out %gpu_0_res3_3_branch2b__5_res // size: 200704 | |
90 %gpu_0_res3_3_branch2c_bn__6 = elementadd @out %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res3_2_branch2c_bn__6_res | |
91 %dealloc_gpu_0_res3_2_branch2c_bn__6_res = deallocactivation @out %gpu_0_res3_2_branch2c_bn__6_res // size: 802816 | |
92 %gpu_0_res3_3_branch2c_bn__7 = relu @out %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res3_3_branch2c_bn__6_res | |
93 %gpu_0_res4_0_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 28 x 28>} // size: 401408 // Users: @out 97, @in 96, @out 94 | |
94 %gpu_0_res4_0_branch2a__5 = convolution @out %gpu_0_res4_0_branch2a__5_res, @in %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res4_0_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__5 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
95 %gpu_0_res4_0_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 100, @in 99, @out 96 | |
96 %gpu_0_res4_0_branch2b__5 = convolution @out %gpu_0_res4_0_branch2b__5_res, @in %gpu_0_res4_0_branch2a__5_res, @in %gpu_0_res4_0_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__6 { Kernels: [3, 3], Strides: [2, 2], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
97 %dealloc_gpu_0_res4_0_branch2a__5_res = deallocactivation @out %gpu_0_res4_0_branch2a__5_res // size: 401408 | |
98 %gpu_0_res4_0_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @in 104, @out 99, @in 115, @in 108, @out 106, @out 116, @in 106, @out 104 | |
99 %gpu_0_res4_0_branch2c__6 = convolution @out %gpu_0_res4_0_branch2c_bn__6_res, @in %gpu_0_res4_0_branch2b__5_res, @in %gpu_0_res4_0_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__2 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
100 %dealloc_gpu_0_res4_0_branch2b__5_res = deallocactivation @out %gpu_0_res4_0_branch2b__5_res // size: 100352 | |
101 %gpu_0_res4_0_branch1__4_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @out 105, @in 104, @out 102 | |
102 %gpu_0_res4_0_branch1__4 = convolution @out %gpu_0_res4_0_branch1__4_res, @in %gpu_0_res3_3_branch2c_bn__6_res, @in %gpu_0_res4_0_branch1_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__1 { Kernels: [1, 1], Strides: [2, 2], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
103 %dealloc_gpu_0_res3_3_branch2c_bn__6_res = deallocactivation @out %gpu_0_res3_3_branch2c_bn__6_res // size: 802816 | |
104 %gpu_0_res4_0_branch2c_bn__6 = elementadd @out %gpu_0_res4_0_branch2c_bn__6_res, @in %gpu_0_res4_0_branch2c_bn__6_res, @in %gpu_0_res4_0_branch1__4_res | |
105 %dealloc_gpu_0_res4_0_branch1__4_res = deallocactivation @out %gpu_0_res4_0_branch1__4_res // size: 401408 | |
106 %gpu_0_res4_0_branch2c_bn__7 = relu @out %gpu_0_res4_0_branch2c_bn__6_res, @in %gpu_0_res4_0_branch2c_bn__6_res | |
107 %gpu_0_res4_1_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 111, @in 110, @out 108 | |
108 %gpu_0_res4_1_branch2a__5 = convolution @out %gpu_0_res4_1_branch2a__5_res, @in %gpu_0_res4_0_branch2c_bn__6_res, @in %gpu_0_res4_1_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__7 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
109 %gpu_0_res4_1_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 114, @in 113, @out 110 | |
110 %gpu_0_res4_1_branch2b__5 = convolution @out %gpu_0_res4_1_branch2b__5_res, @in %gpu_0_res4_1_branch2a__5_res, @in %gpu_0_res4_1_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__8 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
111 %dealloc_gpu_0_res4_1_branch2a__5_res = deallocactivation @out %gpu_0_res4_1_branch2a__5_res // size: 100352 | |
112 %gpu_0_res4_2_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @in 115, @out 113, @in 117, @out 115, @in 126, @in 119, @out 117, @in 137, @in 130, @out 128, @out 138, @in 128, @out 126 | |
113 %gpu_0_res4_1_branch2c__6 = convolution @out %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_1_branch2b__5_res, @in %gpu_0_res4_1_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
114 %dealloc_gpu_0_res4_1_branch2b__5_res = deallocactivation @out %gpu_0_res4_1_branch2b__5_res // size: 100352 | |
115 %gpu_0_res4_1_branch2c_bn__6 = elementadd @out %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_0_branch2c_bn__6_res | |
116 %dealloc_gpu_0_res4_0_branch2c_bn__6_res = deallocactivation @out %gpu_0_res4_0_branch2c_bn__6_res // size: 401408 | |
117 %gpu_0_res4_1_branch2c_bn__7 = relu @out %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2c_bn__6_res | |
118 %gpu_0_res4_2_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 122, @in 121, @out 119 | |
119 %gpu_0_res4_2_branch2a__5 = convolution @out %gpu_0_res4_2_branch2a__5_res, @in %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__9 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
120 %gpu_0_res4_2_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 125, @in 124, @out 121 | |
121 %gpu_0_res4_2_branch2b__5 = convolution @out %gpu_0_res4_2_branch2b__5_res, @in %gpu_0_res4_2_branch2a__5_res, @in %gpu_0_res4_2_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__10 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
122 %dealloc_gpu_0_res4_2_branch2a__5_res = deallocactivation @out %gpu_0_res4_2_branch2a__5_res // size: 100352 | |
123 %gpu_0_res4_2_branch2c__6_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @out 127, @in 126, @out 124 | |
124 %gpu_0_res4_2_branch2c__6 = convolution @out %gpu_0_res4_2_branch2c__6_res, @in %gpu_0_res4_2_branch2b__5_res, @in %gpu_0_res4_2_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__4 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
125 %dealloc_gpu_0_res4_2_branch2b__5_res = deallocactivation @out %gpu_0_res4_2_branch2b__5_res // size: 100352 | |
126 %gpu_0_res4_2_branch2c_bn__6 = elementadd @out %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2c__6_res, @in %gpu_0_res4_2_branch2c_bn__6_res | |
127 %dealloc_gpu_0_res4_2_branch2c__6_res = deallocactivation @out %gpu_0_res4_2_branch2c__6_res // size: 401408 | |
128 %gpu_0_res4_2_branch2c_bn__7 = relu @out %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2c_bn__6_res | |
129 %gpu_0_res4_3_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 133, @in 132, @out 130 | |
130 %gpu_0_res4_3_branch2a__5 = convolution @out %gpu_0_res4_3_branch2a__5_res, @in %gpu_0_res4_2_branch2c_bn__6_res, @in %gpu_0_res4_3_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__11 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
131 %gpu_0_res4_3_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 136, @in 135, @out 132 | |
132 %gpu_0_res4_3_branch2b__5 = convolution @out %gpu_0_res4_3_branch2b__5_res, @in %gpu_0_res4_3_branch2a__5_res, @in %gpu_0_res4_3_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__12 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
133 %dealloc_gpu_0_res4_3_branch2a__5_res = deallocactivation @out %gpu_0_res4_3_branch2a__5_res // size: 100352 | |
134 %gpu_0_res4_3_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @in 137, @out 135, @in 148, @in 141, @out 139, @out 149, @in 139, @out 137 | |
135 %gpu_0_res4_3_branch2c__6 = convolution @out %gpu_0_res4_3_branch2c_bn__6_res, @in %gpu_0_res4_3_branch2b__5_res, @in %gpu_0_res4_3_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__5 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
136 %dealloc_gpu_0_res4_3_branch2b__5_res = deallocactivation @out %gpu_0_res4_3_branch2b__5_res // size: 100352 | |
137 %gpu_0_res4_3_branch2c_bn__6 = elementadd @out %gpu_0_res4_3_branch2c_bn__6_res, @in %gpu_0_res4_3_branch2c_bn__6_res, @in %gpu_0_res4_2_branch2c_bn__6_res | |
138 %dealloc_gpu_0_res4_2_branch2c_bn__6_res = deallocactivation @out %gpu_0_res4_2_branch2c_bn__6_res // size: 401408 | |
139 %gpu_0_res4_3_branch2c_bn__7 = relu @out %gpu_0_res4_3_branch2c_bn__6_res, @in %gpu_0_res4_3_branch2c_bn__6_res | |
140 %gpu_0_res4_4_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 144, @in 143, @out 141 | |
141 %gpu_0_res4_4_branch2a__5 = convolution @out %gpu_0_res4_4_branch2a__5_res, @in %gpu_0_res4_3_branch2c_bn__6_res, @in %gpu_0_res4_4_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__13 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
142 %gpu_0_res4_4_branch2b__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 147, @in 146, @out 143 | |
143 %gpu_0_res4_4_branch2b__5 = convolution @out %gpu_0_res4_4_branch2b__5_res, @in %gpu_0_res4_4_branch2a__5_res, @in %gpu_0_res4_4_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__14 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
144 %dealloc_gpu_0_res4_4_branch2a__5_res = deallocactivation @out %gpu_0_res4_4_branch2a__5_res // size: 100352 | |
145 %gpu_0_res4_4_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @in 148, @out 146, @in 159, @in 152, @out 150, @out 160, @in 150, @out 148 | |
146 %gpu_0_res4_4_branch2c__6 = convolution @out %gpu_0_res4_4_branch2c_bn__6_res, @in %gpu_0_res4_4_branch2b__5_res, @in %gpu_0_res4_4_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold__6 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
147 %dealloc_gpu_0_res4_4_branch2b__5_res = deallocactivation @out %gpu_0_res4_4_branch2b__5_res // size: 100352 | |
148 %gpu_0_res4_4_branch2c_bn__6 = elementadd @out %gpu_0_res4_4_branch2c_bn__6_res, @in %gpu_0_res4_4_branch2c_bn__6_res, @in %gpu_0_res4_3_branch2c_bn__6_res | |
149 %dealloc_gpu_0_res4_3_branch2c_bn__6_res = deallocactivation @out %gpu_0_res4_3_branch2c_bn__6_res // size: 401408 | |
150 %gpu_0_res4_4_branch2c_bn__7 = relu @out %gpu_0_res4_4_branch2c_bn__6_res, @in %gpu_0_res4_4_branch2c_bn__6_res | |
151 %gpu_0_res4_5_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 155, @in 154, @out 152 | |
152 %gpu_0_res4_5_branch2a__5 = convolution @out %gpu_0_res4_5_branch2a__5_res, @in %gpu_0_res4_4_branch2c_bn__6_res, @in %gpu_0_res4_5_branch2a_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold__15 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
153 %gpu_0_res4_5_branch2b__5_res = allocactivation { Ty: i16[S:0.000976562 O:0][-32.000,31.999]<1 x 256 x 14 x 14>} // size: 100352 // Users: @out 158, @in 157, @out 154 | |
154 %gpu_0_res4_5_branch2b__5 = convolution @out %gpu_0_res4_5_branch2b__5_res, @in %gpu_0_res4_5_branch2a__5_res, @in %gpu_0_res4_5_branch2b_w__2, @in %gpu_0_res2_0_branch2c_bias_constfold { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
155 %dealloc_gpu_0_res4_5_branch2a__5_res = deallocactivation @out %gpu_0_res4_5_branch2a__5_res // size: 100352 | |
156 %gpu_0_res4_5_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1024 x 14 x 14>} // size: 401408 // Users: @in 159, @out 157, @in 171, @in 163, @out 161, @out 172, @in 161, @out 159 | |
157 %gpu_0_res4_5_branch2c__6 = convolution @out %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res4_5_branch2b__5_res, @in %gpu_0_res4_5_branch2c_w__2, @in %gpu_0_res4_0_branch2c_bias_constfold { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
158 %dealloc_gpu_0_res4_5_branch2b__5_res = deallocactivation @out %gpu_0_res4_5_branch2b__5_res // size: 100352 | |
159 %gpu_0_res4_5_branch2c_bn__6 = elementadd @out %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res4_4_branch2c_bn__6_res | |
160 %dealloc_gpu_0_res4_4_branch2c_bn__6_res = deallocactivation @out %gpu_0_res4_4_branch2c_bn__6_res // size: 401408 | |
161 %gpu_0_res4_5_branch2c_bn__7 = relu @out %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res4_5_branch2c_bn__6_res | |
162 %gpu_0_res5_0_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 512 x 14 x 14>} // size: 200704 // Users: @out 166, @in 165, @out 163 | |
163 %gpu_0_res5_0_branch2a__5 = convolution @out %gpu_0_res5_0_branch2a__5_res, @in %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res5_0_branch2a_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__6 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
164 %gpu_0_res5_0_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 512 x 7 x 7>} // size: 50176 // Users: @out 169, @in 168, @out 165 | |
165 %gpu_0_res5_0_branch2b__5 = convolution @out %gpu_0_res5_0_branch2b__5_res, @in %gpu_0_res5_0_branch2a__5_res, @in %gpu_0_res5_0_branch2b_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__7 { Kernels: [3, 3], Strides: [2, 2], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
166 %dealloc_gpu_0_res5_0_branch2a__5_res = deallocactivation @out %gpu_0_res5_0_branch2a__5_res // size: 200704 | |
167 %gpu_0_res5_0_branch2c__6_res = allocactivation { Ty: i16[S:0.000976562 O:0][-32.000,31.999]<1 x 2048 x 7 x 7>} // size: 200704 // Users: @out 175, @in 174, @out 168 | |
168 %gpu_0_res5_0_branch2c__6 = convolution @out %gpu_0_res5_0_branch2c__6_res, @in %gpu_0_res5_0_branch2b__5_res, @in %gpu_0_res5_0_branch2c_w__2, @in %gpu_0_res5_0_branch2c_bias_constfold__2 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
169 %dealloc_gpu_0_res5_0_branch2b__5_res = deallocactivation @out %gpu_0_res5_0_branch2b__5_res // size: 50176 | |
170 %gpu_0_res5_0_branch1__4_res = allocactivation { Ty: i16[S:0.000976562 O:0][-32.000,31.999]<1 x 2048 x 7 x 7>} // size: 200704 // Users: @out 176, @in 174, @out 171 | |
171 %gpu_0_res5_0_branch1__4 = convolution @out %gpu_0_res5_0_branch1__4_res, @in %gpu_0_res4_5_branch2c_bn__6_res, @in %gpu_0_res5_0_branch1_w__2, @in %gpu_0_res5_0_branch2c_bias_constfold__1 { Kernels: [1, 1], Strides: [2, 2], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
172 %dealloc_gpu_0_res4_5_branch2c_bn__6_res = deallocactivation @out %gpu_0_res4_5_branch2c_bn__6_res // size: 401408 | |
173 %gpu_0_res5_2_branch2c_bn__6_res = allocactivation { Ty: i16[S:0.001953125 O:0][-64.000,63.998]<1 x 2048 x 7 x 7>} // size: 200704 // Users: @in 177, @out 174, @in 186, @in 179, @out 177, @in 188, @out 186, @in 197, @in 190, @out 188, @in 201, @out 199, @out 202, @in 199, @out 197 | |
174 %gpu_0_res5_0_branch2c_bn__6 = elementadd @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_0_branch2c__6_res, @in %gpu_0_res5_0_branch1__4_res | |
175 %dealloc_gpu_0_res5_0_branch2c__6_res = deallocactivation @out %gpu_0_res5_0_branch2c__6_res // size: 200704 | |
176 %dealloc_gpu_0_res5_0_branch1__4_res = deallocactivation @out %gpu_0_res5_0_branch1__4_res // size: 200704 | |
177 %gpu_0_res5_0_branch2c_bn__7 = relu @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_2_branch2c_bn__6_res | |
178 %gpu_0_res5_1_branch2a__5_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 512 x 7 x 7>} // size: 50176 // Users: @out 182, @in 181, @out 179 | |
179 %gpu_0_res5_1_branch2a__5 = convolution @out %gpu_0_res5_1_branch2a__5_res, @in %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_1_branch2a_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__8 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
180 %gpu_0_res5_1_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 512 x 7 x 7>} // size: 50176 // Users: @out 185, @in 184, @out 181 | |
181 %gpu_0_res5_1_branch2b__5 = convolution @out %gpu_0_res5_1_branch2b__5_res, @in %gpu_0_res5_1_branch2a__5_res, @in %gpu_0_res5_1_branch2b_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__9 { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
182 %dealloc_gpu_0_res5_1_branch2a__5_res = deallocactivation @out %gpu_0_res5_1_branch2a__5_res // size: 50176 | |
183 %gpu_0_res5_1_branch2c__6_res = allocactivation { Ty: i16[S:0.000976562 O:0][-32.000,31.999]<1 x 2048 x 7 x 7>} // size: 200704 // Users: @out 187, @in 186, @out 184 | |
184 %gpu_0_res5_1_branch2c__6 = convolution @out %gpu_0_res5_1_branch2c__6_res, @in %gpu_0_res5_1_branch2b__5_res, @in %gpu_0_res5_1_branch2c_w__2, @in %gpu_0_res5_0_branch2c_bias_constfold__3 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
185 %dealloc_gpu_0_res5_1_branch2b__5_res = deallocactivation @out %gpu_0_res5_1_branch2b__5_res // size: 50176 | |
186 %gpu_0_res5_1_branch2c_bn__6 = elementadd @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_1_branch2c__6_res, @in %gpu_0_res5_2_branch2c_bn__6_res | |
187 %dealloc_gpu_0_res5_1_branch2c__6_res = deallocactivation @out %gpu_0_res5_1_branch2c__6_res // size: 200704 | |
188 %gpu_0_res5_1_branch2c_bn__7 = relu @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_2_branch2c_bn__6_res | |
189 %gpu_0_res5_2_branch2a__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 512 x 7 x 7>} // size: 50176 // Users: @out 193, @in 192, @out 190 | |
190 %gpu_0_res5_2_branch2a__5 = convolution @out %gpu_0_res5_2_branch2a__5_res, @in %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_2_branch2a_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold__10 { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
191 %gpu_0_res5_2_branch2b__5_res = allocactivation { Ty: i16[S:0.000244141 O:0][-8.000,8.000]<1 x 512 x 7 x 7>} // size: 50176 // Users: @out 196, @in 195, @out 192 | |
192 %gpu_0_res5_2_branch2b__5 = convolution @out %gpu_0_res5_2_branch2b__5_res, @in %gpu_0_res5_2_branch2a__5_res, @in %gpu_0_res5_2_branch2b_w__2, @in %gpu_0_res3_0_branch2c_bias_constfold { Kernels: [3, 3], Strides: [1, 1], Pads: [1, 1, 1, 1], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: RELU, FusedActivationArgs: []} | |
193 %dealloc_gpu_0_res5_2_branch2a__5_res = deallocactivation @out %gpu_0_res5_2_branch2a__5_res // size: 50176 | |
194 %gpu_0_res5_2_branch2c__6_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 2048 x 7 x 7>} // size: 200704 // Users: @out 198, @in 197, @out 195 | |
195 %gpu_0_res5_2_branch2c__6 = convolution @out %gpu_0_res5_2_branch2c__6_res, @in %gpu_0_res5_2_branch2b__5_res, @in %gpu_0_res5_2_branch2c_w__2, @in %gpu_0_res5_0_branch2c_bias_constfold { Kernels: [1, 1], Strides: [1, 1], Pads: [0, 0, 0, 0], Group: 1, Dilation: [1, 1], Layout: NCHW, FusedActivation: NONE, FusedActivationArgs: []} | |
196 %dealloc_gpu_0_res5_2_branch2b__5_res = deallocactivation @out %gpu_0_res5_2_branch2b__5_res // size: 50176 | |
197 %gpu_0_res5_2_branch2c_bn__6 = elementadd @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_2_branch2c__6_res, @in %gpu_0_res5_2_branch2c_bn__6_res | |
198 %dealloc_gpu_0_res5_2_branch2c__6_res = deallocactivation @out %gpu_0_res5_2_branch2c__6_res // size: 200704 | |
199 %gpu_0_res5_2_branch2c_bn__7 = relu @out %gpu_0_res5_2_branch2c_bn__6_res, @in %gpu_0_res5_2_branch2c_bn__6_res | |
200 %gpu_0_pool5__4_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 2048 x 1 x 1>} // size: 4096 // Users: @in 203, @out 206, @out 201 | |
201 %gpu_0_pool5__4 = avgpool @out %gpu_0_pool5__4_res, @in %gpu_0_res5_2_branch2c_bn__6_res { Kernels: [7, 7], Strides: [1, 1], Pads: [0, 0, 0, 0], Layout: 1, CountIncludePads: 1} | |
202 %dealloc_gpu_0_res5_2_branch2c_bn__6_res = deallocactivation @out %gpu_0_res5_2_branch2c_bn__6_res // size: 200704 | |
203 %gpu_0_pool5__4_res__2 = tensorview @in %gpu_0_pool5__4_res { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 2048>, Offsets: [0, 0, 0, 0]} // Users: @in 205 | |
204 %gpu_0_pred_res = allocactivation { Ty: i16[S:0.000488281 O:0][-16.000,16.000]<1 x 1000>} // size: 2000 // Users: @out 209, @in 208, @out 205 | |
205 %gpu_0_pred = fullyconnected @out %gpu_0_pred_res, @in %gpu_0_pool5__4_res__2, @in %gpu_0_pred_w__2, @in %gpu_0_pred_b | |
206 %dealloc_gpu_0_pool5__4_res = deallocactivation @out %gpu_0_pool5__4_res // size: 4096 | |
207 %gpu_0_softmax__1_res = allocactivation { Ty: i16[S:0.000030518 O:0][-1.000,1.000]<1 x 1000>} // size: 2000 // Users: @out 211, @in 210, @out 208 | |
208 %gpu_0_softmax__1 = softmax @out %gpu_0_softmax__1_res, @in %gpu_0_pred_res | |
209 %dealloc_gpu_0_pred_res = deallocactivation @out %gpu_0_pred_res // size: 2000 | |
210 %gpu_0_softmax__1_dequantize = dequantize @out %gpu_0_softmax, @in %gpu_0_softmax__1_res | |
211 %dealloc_gpu_0_softmax__1_res = deallocactivation @out %gpu_0_softmax__1_res // size: 2000 | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment