IgorMirosavljevicHTEC
commented
6 months ago | model | default rate |
extended rate |
perf diff |
time diff |
|---|---|---|---|---|
| torchvision-resnet50 | 21.4707 | 21.0699 | -0.41297 :white_check_mark: | 27m28s |
| torchvision-resnet50_fp16 | 9.405 | 9.05957 | -0.35678 :white_check_mark: | 19m18s |
| torchvision-densenet121 | 13.3029 | 13.1682 | -0.13685 :white_check_mark: | 22m57s |
| torchvision-densenet121_fp16 | 7.8712 | 7.69926 | -0.1284 :white_check_mark: | 15m35s |
| torchvision-inceptionv3 | 19.3873 | 19.2296 | -0.16688 :white_check_mark: | 11m5s |
| torchvision-inceptionv3_fp16 | 12.2936 | 11.9804 | -0.32055 :white_check_mark: | 12m32s |
| cadene-inceptionv4 | 20.7222 | 19.9217 | -0.81714 :white_check_mark: | 7m50s |
| cadene-resnext64x4 | 21.7808 | 21.0924 | -0.65743 :white_check_mark: | 12m35s |
| slim-mobilenet | 9.17152 | 7.68672 | -1.48949 :white_check_mark: | 25m1s |
| slim-nasnetalarge | 362.792 | 319.387 | -43.9405 :white_check_mark: | 114m7s |
| slim-resnet50v2 | 22.4708 | 21.7051 | -0.78866 :white_check_mark: | 21m31s |
| bert-mrpc-onnx | 7.66774 | 7.19975 | -0.47715 :white_check_mark: | 2m5s |
| bert-mrpc-tf | 2.12468 | 2.39772 | 0.148775 :red_circle: | 1m36s |
| pytorch-examples-wlang-gru | 1.9022 | 2.25884 | 0.149624 :red_circle: | 0m52s |
| pytorch-examples-wlang-lstm | 2.10547 | 2.57344 | 0.093087 :red_circle: | 0m46s |
| torchvision-resnet50_1 | 1.36961 | 1.27107 | 0.05001 :red_circle: | 2m35s |
| cadene-dpn92_1 | 2.84821 | 2.85613 | 0.324351 :red_circle: | 6m48s |
| cadene-resnext101_1 | 2.75558 | 2.68467 | -0.05946 :white_check_mark: | 3m13s |
| onnx-taau-downsample | 2.85933 | 2.83414 | -0.01402 :white_check_mark: | 2m7s |
| dlrm-criteoterabyte | 30.0065 | 28.1084 | -2.00345 :white_check_mark: | 85m21s |
| dlrm-criteoterabyte_fp16 | 17.1232 | 16.8194 | -0.21323 :white_check_mark: | 57m12s |
| agentmodel | 0.138447 | 0.161597 | -0.0421 :white_check_mark: | 0m20s |
| unet_fp16 | 33.9634 | 31.8342 | -2.58295 :white_check_mark: | 21m37s |
| resnet50v1_fp16 | 1.03229 | 0.979984 | -0.06999 :white_check_mark: | 8m57s |
| resnet50v1_int8 | 1.2569 | 1.02615 | -0.17043 :white_check_mark: | 23m29s |
| bert_base_cased_fp16 | 64.1911 | 61.6807 | -3.04018 :white_check_mark: | 17m29s |
| bert_large_uncased_fp16 | 106.162 | 100.319 | -5.86713 :white_check_mark: | 12m42s |
| bert_large_fp16 | 5.10241 | 4.70123 | -0.34 :white_check_mark: | 7m50s |
| distilgpt2_fp16 | 8.03759 | 7.86348 | -0.18677 :white_check_mark: | 4m31s |
| yolov5s | 1.90073 | 1.86153 | -0.07495 :white_check_mark: | 1m53s |
| tinyllama | 22.3574 | 19.357 | -3.04506 :white_check_mark: | 3m4s |
| vicuna-fastchat | 5.41466 | 5.39066 | -0.06952 :white_check_mark: | 1m29s |
| whisper-tiny-encoder | 2.4532 | 2.31435 | -0.16052 :white_check_mark: | 2m20s |
| whisper-tiny-decoder | 2.38854 | 2.12036 | -0.27047 :white_check_mark: | 3m56s |
Extended performance test with "exhaustive-tune" and "MIGRAPHX_MLIR_USE_SPECIFIC_OPS" set