pytorch/benchmark

Pull Requests Commits

Reduce repeat/warmup for faster results

wconstab committed 3 years ago

98f03af6

set PYTORCH_NVFUSER_DISABLE_FALLBACK=1

wconstab committed 3 years ago

fab9c5c0

fix niter kwarg for moco, stargan, cyclegan

wconstab committed 3 years ago

9f3d54a7

Merge branch 'wconstab/ltc' of github.com:pytorch/benchmark into wconstab/ltc

wconstab committed 4 years ago

c6579cb4

[wconstab/ltc] Skip timm_nfnet training (#670)

alanwaketan committed 4 years ago

Verified fa0e0ac0

[wconstab/ltc] Skip running measurement for the base commit (#662)

alanwaketan committed 4 years ago

Verified 7efef87f

[wconstab/ltc] Skip dlrm training test (#655)

alanwaketan committed 4 years ago

Verified 81e038a3

[wconstab/ltc] Enable lazy_bench.py for the TB CI (#632)

alanwaketan committed 4 years ago

Verified c3258965

Double the time out limits

Jiewen Tan committed 4 years ago

9f157d63

Remove torch, torchvision requirements from models since that can cause overwriting the dev/CI env's torch

wconstab committed 4 years ago

acda8759

Merge branch 'main' into wconstab/ltc

Jiewen Tan committed 4 years ago

a0e0d459

Make the script more robust on supporting unknown devices (#642)

Jiewen Tan committed 4 years ago

79798c58

Make the model robust to accept Lazy device (#646)

Jiewen Tan committed 4 years ago

c2b7cacc

[wconstab/ltc] Add in-memory test expectations (#647)

alanwaketan committed 4 years ago

Verified d36e73e7

[wconstab/ltc] Make nvidia_deeprecommender accept lazy device (#643)

alanwaketan committed 4 years ago

Verified ad21edb9

Unconditionally move inputs to the specified device (#637)

Jiewen Tan committed 4 years ago

8f5048a4

Avoid instantiating models for the sake of checking their names (#641)

Krovatkin committed 4 years ago

Verified 177d67ce

Fix yolov3 model training to original batch size (#630)

xuzhao9 committed 4 years ago

0954aa61

DCGan Sample As Benchmark (#448)

eircfb committed 4 years ago

df9aa52c

Fix timm_nfnet train batch size and improve the entire model quality (#585)

xuzhao9 committed 4 years ago

848d8dfd

Handle timeouts w/o halting the whole sweep and master script (#638)

Krovatkin committed 4 years ago

Verified c1111291

Remove two wav files in tacotron2. (#633)

xuzhao9 committed 4 years ago

74247aac

Allow install.py to only install deps of specified models. (#631)

xuzhao9 committed 4 years ago

385fc6ff

dont format time (#636)

Krovatkin committed 4 years ago

Verified ae6914a9

Added v2 workflows and prepare for the test run of the benchmark (#624)

xuzhao9 committed 4 years ago

0de79002

Merge branch 'main' into wconstab/ltc

Jiewen Tan committed 4 years ago

ac824d9f

[wconstab/ltc] Convert the json output from check_lazy.py to csv (#623)

alanwaketan committed 4 years ago

Verified 50b56a1a

add timestamps to model launches (#619)

Krovatkin committed 4 years ago

Verified 15d22c6c

Fix several bugs in the bisection script. (#628)

xuzhao9 committed 4 years ago

8b6833fa

Fix training hparams in Tacotron2 (#610)

aaronenyeshi committed 4 years ago

9ade7265

Older