Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 20, 2025

• add arguments to the respective argument groups

  

  JosuaRieder

  authored and

  rwightman

  committed
  cb4cea5

  Copy full SHA for cb4cea5

• Fix metavar for --input-size

  

  JosuaRieder

  authored and

  rwightman

  committed
  634b68a

  Copy full SHA for 634b68a

Commits on Jan 9, 2025

• Avoid unecessary compat break btw train script and nearby timm versions w/ dtype addition.

  rwightman

  committed
  eeee38e

  Copy full SHA for eeee38e

Commits on Jan 8, 2025

• Merge branch 'main' into caojiaolong-main

  rwightman

  committed
  c173886

  Copy full SHA for c173886

Commits on Jan 7, 2025

• Fix dtype log when default (None) is used w/o AMP

  rwightman

  committed
  1969528

  Copy full SHA for 1969528

• Add half-precision (bfloat16, float16) support to train & validate scripts. Should push dtype handling into model factory / pretrained load at some point...

  rwightman

  committed
  92f610c

  Copy full SHA for 92f610c

• Add wandb project name argument and allow change wandb run name

  caojiaolong

  authored
  40c19f3

  Copy full SHA for 40c19f3

Commits on Dec 6, 2024

• Merge branch 'main' of github.com:grodino/pytorch-image-models into grodino-dataset_trust_remote

  rwightman

  committed
  95d903f

  Copy full SHA for 95d903f

Commits on Nov 21, 2024

• Improve wandb logging

  

  sinahmr

  authored and

  rwightman

  committed
  3a6cc4f

  Copy full SHA for 3a6cc4f

• Improve the parsable results dump at end of train, stop excessive output, only display top-10.

  rwightman

  committed
  620cb4f

  Copy full SHA for 620cb4f

• In dist training, update loss running avg every step, only sync on log updates / final.

  rwightman

  committed
  36b5d1a

  Copy full SHA for 36b5d1a

Commits on Nov 13, 2024

• A bit of an optimizer overhaul, added an improved factory, list_optimizers, class helper and add info classes with descriptions, arg configs

  rwightman

  committed
  ee5f6e7

  Copy full SHA for ee5f6e7

Commits on Nov 8, 2024

• Update log to describe scheduling behaviour diff w/ warmup_prefix

  rwightman

  committed
  9f5c279

  Copy full SHA for 9f5c279

• Add --dataset-trust-remote-code to the train.py and validate.py scripts

  grodino

  committed
  2dff16f

  Copy full SHA for 2dff16f

Commits on Nov 5, 2024

• fix typo in train.py: bathes > batches

  

  JosuaRieder

  authored and

  rwightman

  committed
  51ac8d2

  Copy full SHA for 51ac8d2

Commits on Oct 18, 2024

• Merge branch 'MengqingCao-npu_support' into device_amp_cleanup

  rwightman

  committed
  015fbe4

  Copy full SHA for 015fbe4

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

  rwightman

  committed
  1766a01

  Copy full SHA for 1766a01

Commits on Oct 16, 2024

• add npu support

  MengqingCao

  committed
  234f975

  Copy full SHA for 234f975

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 Jul 22, 2024

• Fix LR scheduler help in train.py

  Show description for 4ed93fc
  Zirunis

  authored
  4ed93fc

  Copy full SHA for 4ed93fc

Commits on Jul 15, 2024

• Avoid zero division error

  TianyiFranklinWang

  committed
  d3ce5a8

  Copy full SHA for d3ce5a8

Commits on Apr 10, 2024

• Fix #2097 a small typo in train.py

  rwightman

  committed
  e25bbfc

  Copy full SHA for e25bbfc

Commits on Feb 10, 2024

• Remove test MESA support, no signal that it's helpful so far

  rwightman

  committed
  5a58f4d

  Copy full SHA for 5a58f4d

Commits on Feb 4, 2024

• Add device arg to validate() calls in train.py

  rwightman

  committed
  c7ac376

  Copy full SHA for c7ac376

• forward() pass through for ema model, flag for ema warmup, comment about warmup

  rwightman

  committed
  bee0471

  Copy full SHA for bee0471

Commits on Feb 2, 2024

• Merge branch 'device_flex' into mesa_ema

  rwightman

  committed
  5e4a4b2

  Copy full SHA for 5e4a4b2

• ModelEmaV3 and MESA experiments

  rwightman

  committed
  dd84ef2

  Copy full SHA for dd84ef2

Commits on Jan 25, 2024

• Pass train-crop-mode to create_loader/transforms from train.py args

  rwightman

  committed
  809a9e1

  Copy full SHA for 809a9e1

Commits on Jan 20, 2024

• Improving device flexibility in train. Fix #2081

  rwightman

  committed
  a48ab81

  Copy full SHA for a48ab81

Commits on Jan 13, 2024

• Fixed index out of range in case of resume

  

  lorenzbaraldi

  authored and

  rwightman

  committed
  8c663c4

  Copy full SHA for 8c663c4

Commits on Jan 8, 2024

• Allow training w/o validation split set

  rwightman

  committed
  c50004d

  Copy full SHA for c50004d

• Significant transforms, dataset, dataloading enhancements.

  rwightman

  committed
  be0944e

  Copy full SHA for be0944e

Commits on Dec 30, 2023

• Add pos_weight and support for summing over classes to BCE impl in train scripts

  rwightman

  committed
  b5a4fa9

  Copy full SHA for b5a4fa9

Commits on Dec 22, 2023

• Add parsable json results output for train.py, tweak --pretrained-path to force head adaptation

  rwightman

  committed
  f2fdd97

  Copy full SHA for f2fdd97

Commits on Dec 11, 2023

• Add --pretrained-path arg to train script to allow passing local checkpoint as pretrained. Add missing/unexpected keys log.

  rwightman

  committed
  60b170b

  Copy full SHA for 60b170b

Pagination

Previous

Next