Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 20, 2025

• Fix metavar for --input-size

  

  JosuaRieder

  authored and

  rwightman

  committed
  634b68a

  Copy full SHA for 634b68a

Commits on Oct 18, 2024

• Cleanup some amp related behaviour to better support different (non-cuda) devices

  rwightman

  committed
  1766a01

  Copy full SHA for 1766a01

Commits on Oct 2, 2024

• Add --torchcompile-mode args to train, validation, inference, benchmark scripts

  rwightman

  committed
  4d4bdd6

  Copy full SHA for 4d4bdd6

Commits on Aug 24, 2023

• MobileOne and FastViT weights on HF hub, more code cleanup and tweaks, features_only working. Add reparam flag to validate and benchmark, support reparm of all models with fuse(), reparameterize() …

  Show description for 5242ba6
  rwightman

  committed
  5242ba6

  Copy full SHA for 5242ba6

Commits on Jan 15, 2023

• Davit update formatting and fix grad checkpointing (#7)

  Show description for 81ca323
  fffffgggg54

  authored
  81ca323

  Copy full SHA for 81ca323

Commits on Dec 9, 2022

• Merge remote-tracking branch 'origin/main' into refactor-imports

  rwightman

  committed
  d5e7d6b

  Copy full SHA for d5e7d6b

• Add a deprecation phase to module re-org

  rwightman

  committed
  cda39b3

  Copy full SHA for cda39b3

Commits on Dec 7, 2022

• Add EVA FT results, hopefully fix BEiT test failures

  rwightman

  committed
  98047ef

  Copy full SHA for 98047ef

Commits on Dec 5, 2022

• Update scripts to support torch.compile(). Make --results_file arg more consistent across benchmark/validate/inference. Fix #1570

  rwightman

  committed
  dbe7531

  Copy full SHA for dbe7531

• Add crop_mode for pretraind config / image transforms. Add support for dynamo compilation to benchmark/train/validate

  rwightman

  committed
  9da7e3a

  Copy full SHA for 9da7e3a

Commits on Sep 23, 2022

• Refactor device handling in scripts, distributed init to be less 'cuda' centric. More device args passed through where needed.

  rwightman

  committed
  87939e6

  Copy full SHA for 87939e6

Commits on Aug 26, 2022

• Add --fast-norm arg to benchmark.py, train.py, validate.py

  rwightman

  committed
  ff6a919

  Copy full SHA for ff6a919

Commits on Jul 19, 2022

• Add bulk_runner script and updates to benchmark.py and validate.py for better error handling in bulk runs (used for benchmark and validation result runs). Improved batch size decay stepping on retr…

  Show description for 0dbd935
  rwightman

  committed
  0dbd935

  Copy full SHA for 0dbd935

Commits on Jul 7, 2022

• Reorg benchmark.py import

  rwightman

  committed
  4670d37

  Copy full SHA for 4670d37

• Add --no-retry flag to benchmark.py to skip batch_size decay and retry on error. Fix #1226. Update deepspeed profile usage for latest DS releases. Fix # 1333

  rwightman

  committed
  28e0152

  Copy full SHA for 28e0152

Commits on Jul 1, 2022

• move dataconfig before script, scripting killing metadata now (PyTorch 1.12? just nvfuser?)

  rwightman

  committed
  34f382f

  Copy full SHA for 34f382f

Commits on Jun 9, 2022

• Move aot-autograd opt after model metadata used to setup data config in benchmark.py

  rwightman

  committed
  2d7ab06

  Copy full SHA for 2d7ab06

Commits on Jun 8, 2022

• add --aot-autograd

  xwang233

  committed
  ca991c1

  Copy full SHA for ca991c1

Commits on Feb 28, 2022

• Significant model refactor and additions:

  Show description for 372ad5f
  rwightman

  committed
  372ad5f

  Copy full SHA for 372ad5f

Commits on Jan 26, 2022

• Merge remote-tracking branch 'origin/master' into norm_norm_norm

  rwightman

  committed
  95cfc9b

  Copy full SHA for 95cfc9b

Commits on Jan 24, 2022

• Update benchmark and validate scripts to output results in JSON with a fixed delimiter for use in multi-process launcher

  rwightman

  committed
  cf43343

  Copy full SHA for cf43343

Commits on Jan 21, 2022

• refactor: remove if-condition

  kozistr

  committed
  56a6b38

  Copy full SHA for 56a6b38

Commits on Jan 17, 2022

• Add --fuser arg to train/validate/benchmark scripts to select jit fuser type

  rwightman

  committed
  f0f9ecc

  Copy full SHA for f0f9ecc

Commits on Dec 14, 2021

• Add drop args to benchmark.py

  rwightman

  committed
  683fba7

  Copy full SHA for 683fba7

Commits on Oct 28, 2021

• Add new 50ts attn models to benchmark/meta csv files

  rwightman

  committed
  aaff2d8

  Copy full SHA for aaff2d8

• Fixed botne*t26 model results, add some 50ts self-attn variants

  rwightman

  committed
  1e17863

  Copy full SHA for 1e17863

Commits on Oct 25, 2021

• Don't run profile if model is torchscripted

  rwightman

  committed
  71f00bf

  Copy full SHA for 71f00bf

Commits on Oct 24, 2021

• Add activation count to fvcore based profiling in benchmark.py

  rwightman

  committed
  5882e62

  Copy full SHA for 5882e62

Commits on Oct 20, 2021

• Support either deepspeed or fvcore for flop profiling

  rwightman

  committed
  f7325c7

  Copy full SHA for f7325c7

Commits on Oct 19, 2021

• Add `--bench profile` mode for benchmark.py to just run deepspeed detailed profile on model

  rwightman

  committed
  6625379

  Copy full SHA for 6625379

• Add train size override and deepspeed GMACs counter (if deepspeed installed) to benchmark.py

  rwightman

  committed
  13a8bf7

  Copy full SHA for 13a8bf7

Commits on Aug 18, 2021

• Optimizer improvements, additions, cleanup

  Show description for ac469b5
  rwightman

  committed
  ac469b5

  Copy full SHA for ac469b5

Commits on May 4, 2021

• Merge pull request #555 from MichaelMonashev/patch-1

  Show description for 137a374
  rwightman

  authored
  137a374

  Copy full SHA for 137a374

Commits on Apr 16, 2021

• Fix #566, summary.csv writing to pwd on local_rank != 0. Tweak benchmark mem handling to see if it reduces likelihood of 'bad' exceptions on OOM.

  rwightman

  committed
  e15e68d

  Copy full SHA for e15e68d

Commits on Apr 11, 2021

• Argument description fixed

  MichaelMonashev

  authored
  0be1fa4

  Copy full SHA for 0be1fa4

Pagination

Previous

Next