YoloV3 Tiny quantization tuned parameters

yolov3-tiny.tune.log

{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 256, 26, 26], "int8"], ["TENSOR", [255, 256, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 26, 26, "int8"], [255, 256, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 5, "t": "direct", "c": null, "e": [["tile_ic", "sp", [8, 32]], ["tile_oc", "sp", [255, 1]], ["tile_ow", "sp", [26, 1]], ["tile_oh", "ot", 1]]}], "r": [[0.002006969857142857], 0, 1.7788748741149902, 1564778173.0481853], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 384, 26, 26], "int8"], ["TENSOR", [256, 384, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 384, 26, 26, "int8"], [256, 384, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 525, "t": "direct", "c": null, "e": [["tile_ic", "sp", [3, 128]], ["tile_oc", "sp", [8, 32]], ["tile_ow", "sp", [1, 26]], ["unroll_kw", "ot", true]]}], "r": [[0.01963637225], 0, 5.333110809326172, 1564779398.703997], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 256, 13, 13], "int8"], ["TENSOR", [128, 256, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 13, 13, "int8"], [128, 256, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 118, "t": "direct", "c": null, "e": [["tile_ic", "sp", [128, 2]], ["tile_oc", "sp", [4, 32]], ["tile_ow", "sp", [1, 13]], ["tile_oh", "ot", 1]]}], "r": [[0.0002178497583108715], 0, 2.311330556869507, 1564795342.1770709], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 1024, 13, 13], "int8"], ["TENSOR", [256, 1024, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 1024, 13, 13, "int8"], [256, 1024, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 155, "t": "direct", "c": null, "e": [["tile_ic", "sp", [512, 2]], ["tile_oc", "sp", [8, 32]], ["tile_ow", "sp", [1, 13]], ["tile_oh", "ot", 1]]}], "r": [[0.001598656734042553], 0, 1.685750961303711, 1564796026.7763588], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 512, 13, 13], "int8"], ["TENSOR", [1024, 512, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 512, 13, 13, "int8"], [1024, 512, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 178, "t": "direct", "c": null, "e": [["tile_ic", "sp", [2, 256]], ["tile_oc", "sp", [16, 64]], ["tile_ow", "sp", [1, 13]], ["unroll_kw", "ot", true]]}], "r": [[0.02639495795], 0, 5.337036609649658, 1564796641.1192868], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 256, 13, 13], "int8"], ["TENSOR", [512, 256, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 13, 13, "int8"], [512, 256, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 152, "t": "direct", "c": null, "e": [["tile_ic", "sp", [1, 256]], ["tile_oc", "sp", [8, 64]], ["tile_ow", "sp", [1, 13]], ["unroll_kw", "ot", true]]}], "r": [[0.00659780596875], 0, 4.606190919876099, 1564798675.8747852], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 128, 26, 26], "int8"], ["TENSOR", [256, 128, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 128, 26, 26, "int8"], [256, 128, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 261, "t": "direct", "c": null, "e": [["tile_ic", "sp", [4, 32]], ["tile_oc", "sp", [8, 32]], ["tile_ow", "sp", [1, 26]], ["unroll_kw", "ot", true]]}], "r": [[0.00655543884375], 0, 4.97831654548645, 1564799596.0328188], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 64, 52, 52], "int8"], ["TENSOR", [128, 64, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 64, 52, 52, "int8"], [128, 64, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 264, "t": "direct", "c": null, "e": [["tile_ic", "sp", [2, 32]], ["tile_oc", "sp", [4, 32]], ["tile_ow", "sp", [2, 26]], ["unroll_kw", "ot", true]]}], "r": [[0.00657052490625], 0, 5.64388108253479, 1564800946.9031048], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 32, 104, 104], "int8"], ["TENSOR", [64, 32, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 32, 104, 104, "int8"], [64, 32, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 244, "t": "direct", "c": null, "e": [["tile_ic", "sp", [2, 16]], ["tile_oc", "sp", [2, 32]], ["tile_ow", "sp", [4, 26]], ["unroll_kw", "ot", true]]}], "r": [[0.00665656015625], 0, 6.071219205856323, 1564802302.4815693], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 16, 208, 208], "int8"], ["TENSOR", [32, 16, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 16, 208, 208, "int8"], [32, 16, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 174, "t": "direct", "c": null, "e": [["tile_ic", "sp", [1, 16]], ["tile_oc", "sp", [2, 16]], ["tile_ow", "sp", [13, 16]], ["unroll_kw", "ot", true]]}], "r": [[0.00682161653125], 0, 4.296415328979492, 1564803734.1373382], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 3, 416, 416], "float32"], ["TENSOR", [16, 3, 3, 3], "float32"], [1, 1], [1, 1], [1, 1], "NCHW", "float32"], {}, ["conv2d", [1, 3, 416, 416, "float32"], [16, 3, 3, 3, "float32"], [1, 1], [1, 1], [1, 1], "NCHW", "float32"], {"i": 19, "t": "direct", "c": null, "e": [["tile_ic", "sp", [1, 3]], ["tile_oc", "sp", [1, 16]], ["tile_ow", "sp", [208, 2]], ["unroll_kw", "ot", true]]}], "r": [[0.0015622485533333334], 0, 2.444446563720703, 1564804860.316766], "v": 0.1}
{"i": ["llvm", "topi_x86_conv2d_NCHWc", [["TENSOR", [1, 512, 13, 13], "int8"], ["TENSOR", [255, 512, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 512, 13, 13, "int8"], [255, 512, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 5, "t": "direct", "c": null, "e": [["tile_ic", "sp", [16, 32]], ["tile_oc", "sp", [255, 1]], ["tile_ow", "sp", [13, 1]], ["tile_oh", "ot", 1]]}], "r": [[0.0011567560384615385], 0, 1.324392557144165, 1564805611.3352888], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 256, 26, 26], "int8"], ["TENSOR", [255, 256, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 26, 26, "int8"], [255, 256, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 72787, "t": "direct", "c": null, "e": [["tile_f", "sp", [5, 3, 17, 1]], ["tile_y", "sp", [13, 2, 1, 1]], ["tile_x", "sp", [1, 2, 13, 1]], ["tile_rc", "sp", [16, 16]], ["tile_ry", "sp", [1, 1]], ["tile_rx", "sp", [1, 1]], ["auto_unroll_max_step", "ot", 0], ["unroll_explicit", "ot", 0]]}], "r": [[2.8641700495348558e-05], 0, 4.05173397064209, 1564842217.2867043], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 384, 26, 26], "int8"], ["TENSOR", [256, 384, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 384, 26, 26, "int8"], [256, 384, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 7683085, "t": "direct", "c": null, "e": [["tile_f", "sp", [16, 2, 8, 1]], ["tile_y", "sp", [13, 1, 2, 1]], ["tile_x", "sp", [1, 1, 2, 13]], ["tile_rc", "sp", [48, 8]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 0]]}], "r": [[0.00018373059863429438], 0, 4.930304527282715, 1564857950.7440712], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 256, 13, 13], "int8"], ["TENSOR", [128, 256, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 13, 13, "int8"], [128, 256, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 96991, "t": "direct", "c": null, "e": [["tile_f", "sp", [2, 2, 32, 1]], ["tile_y", "sp", [13, 1, 1, 1]], ["tile_x", "sp", [1, 1, 13, 1]], ["tile_rc", "sp", [8, 32]], ["tile_ry", "sp", [1, 1]], ["tile_rx", "sp", [1, 1]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 1]]}], "r": [[9.442641502386387e-06], 0, 6.683902978897095, 1564863272.6469746], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 1024, 13, 13], "int8"], ["TENSOR", [256, 1024, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 1024, 13, 13, "int8"], [256, 1024, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 159751, "t": "direct", "c": null, "e": [["tile_f", "sp", [8, 2, 16, 1]], ["tile_y", "sp", [13, 1, 1, 1]], ["tile_x", "sp", [1, 1, 13, 1]], ["tile_rc", "sp", [32, 32]], ["tile_ry", "sp", [1, 1]], ["tile_rx", "sp", [1, 1]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 1]]}], "r": [[5.722498691410897e-05], 0, 5.194287061691284, 1564866847.4453232], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 512, 13, 13], "int8"], ["TENSOR", [1024, 512, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 512, 13, 13, "int8"], [1024, 512, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 474567, "t": "direct", "c": null, "e": [["tile_f", "sp", [64, 1, 8, 2]], ["tile_y", "sp", [1, 1, 1, 13]], ["tile_x", "sp", [1, 1, 13, 1]], ["tile_rc", "sp", [64, 8]], ["tile_ry", "sp", [3, 1]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 0]]}], "r": [[0.0003258057150537634], 0, 6.138919830322266, 1564877491.7268808], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 256, 13, 13], "int8"], ["TENSOR", [512, 256, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 256, 13, 13, "int8"], [512, 256, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 618505, "t": "direct", "c": null, "e": [["tile_f", "sp", [16, 1, 16, 2]], ["tile_y", "sp", [1, 1, 1, 13]], ["tile_x", "sp", [1, 1, 13, 1]], ["tile_rc", "sp", [16, 16]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 1]]}], "r": [[0.00012775779883536262], 0, 6.941412925720215, 1564883738.4254038], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 128, 26, 26], "int8"], ["TENSOR", [256, 128, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 128, 26, 26, "int8"], [256, 128, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 2505066, "t": "direct", "c": null, "e": [["tile_f", "sp", [4, 2, 32, 1]], ["tile_y", "sp", [1, 1, 2, 13]], ["tile_x", "sp", [13, 1, 2, 1]], ["tile_rc", "sp", [16, 8]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 0]]}], "r": [[5.9136756743380345e-05], 0, 5.322611093521118, 1564891300.9753335], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 64, 52, 52], "int8"], ["TENSOR", [128, 64, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 64, 52, 52, "int8"], [128, 64, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 31539988, "t": "direct", "c": null, "e": [["tile_f", "sp", [2, 4, 16, 1]], ["tile_y", "sp", [2, 1, 2, 13]], ["tile_x", "sp", [13, 1, 4, 1]], ["tile_rc", "sp", [8, 8]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 1]]}], "r": [[4.451105908337935e-05], 0, 8.119948387145996, 1564899019.1694586], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 32, 104, 104], "int8"], ["TENSOR", [64, 32, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 32, 104, 104, "int8"], [64, 32, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 62485379, "t": "direct", "c": null, "e": [["tile_f", "sp", [1, 2, 16, 2]], ["tile_y", "sp", [4, 13, 1, 2]], ["tile_x", "sp", [13, 1, 8, 1]], ["tile_rc", "sp", [8, 4]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 1]]}], "r": [[3.8679101994486784e-05], 0, 8.085409164428711, 1564913091.6077442], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 16, 208, 208], "int8"], ["TENSOR", [32, 16, 3, 3], "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 16, 208, 208, "int8"], [32, 16, 3, 3, "int8"], [1, 1], [1, 1], [1, 1], "NCHW", "int32"], {"i": 129775761, "t": "direct", "c": null, "e": [["tile_f", "sp", [1, 4, 8, 1]], ["tile_y", "sp", [16, 13, 1, 1]], ["tile_x", "sp", [13, 1, 16, 1]], ["tile_rc", "sp", [2, 8]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 1500], ["unroll_explicit", "ot", 1]]}], "r": [[4.5581816399623e-05], 0, 7.952688455581665, 1564921769.5167341], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 3, 416, 416], "float32"], ["TENSOR", [16, 3, 3, 3], "float32"], [1, 1], [1, 1], [1, 1], "NCHW", "float32"], {}, ["conv2d", [1, 3, 416, 416, "float32"], [16, 3, 3, 3, "float32"], [1, 1], [1, 1], [1, 1], "NCHW", "float32"], {"i": 69245459, "t": "direct", "c": null, "e": [["tile_f", "sp", [1, 1, 8, 2]], ["tile_y", "sp", [104, 1, 2, 2]], ["tile_x", "sp", [13, 1, 16, 2]], ["tile_rc", "sp", [1, 3]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [1, 3]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 1]]}], "r": [[4.2642078282382044e-05], 0, 5.567877292633057, 1564939526.0770023], "v": 0.1}
{"i": ["cuda", "topi_nn_conv2d", [["TENSOR", [1, 512, 13, 13], "int8"], ["TENSOR", [255, 512, 1, 1], "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {}, ["conv2d", [1, 512, 13, 13, "int8"], [255, 512, 1, 1, "int8"], [1, 1], [0, 0], [1, 1], "NCHW", "int32"], {"i": 15905, "t": "direct", "c": null, "e": [["tile_f", "sp", [5, 1, 17, 3]], ["tile_y", "sp", [13, 1, 1, 1]], ["tile_x", "sp", [1, 1, 13, 1]], ["tile_rc", "sp", [16, 32]], ["tile_ry", "sp", [1, 1]], ["tile_rx", "sp", [1, 1]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 0]]}], "r": [[3.230471591217132e-05], 0, 5.080807447433472, 1564946497.85989], "v": 0.1}

Hi, I am trying to run yolov3-tiny on pynq-Z2 with VTA. I read about you in the blog of TVM, have you got any result on VTA implementation?

Many thanks
Daniel