Commit Graph

Select branches

Hide Pull Requests

0.5.x

0.6.x

attn_update

bits_and_tpu

cleanup_xla_model_fixes

clip_laion2b

convnext_and_copyright

convnextv2

dataset_info

edgenext_csp_and_more

efficientnet_attn

eva

fix_tests

fix_window_trace

fixes-syncbn_pretrain_cfg_resolve

fixes_bce_regnet

focalnet_and_swin_refactor

freeze_norm_revisit

gh-pages

hf_hub_revisit

levit_efficientformer_redux

main

maxxvit_hub

more_attn

more_datasets

more_vit

multi-weight

multi-weight_effnet_convnext

norm_mlp_classifier

norm_norm_norm

parse_kwargs

parse_kwargs2

perceiver

refactor-imports

scaling_vit

script_cleanup

swin_v2

update-test-pytorch

vit_and_bit_test_fixes

vit_relpos_refactor

xsepconv

#1

#1007

#101

#1012

#1014

#1034

#1037

#105

#1055

#1056

#1056

#1058

#1061

#1063

#1063

#1069

#1083

#1091

#1094

#1105

#1112

#1113

#115

#1150

#117

#1187

#1190

#1194

#12

#1208

#1210

#1210

#1212

#1213

#1213

#122

#1222

#1222

#1229

#123

#1230

#1233

#1239

#1245

#1249

#125

#1251

#1253

#1253

#1256

#1259

#1265

#1266

#1270

#1271

#1271

#1273

#1274

#1276

#1277

#1294

#13

#1317

#132

#1322

#1327

#1330

#1330

#1336

#1340

#1345

#1349

#1351

#1354

#136

#1363

#1365

#137

#1381

#1383

#1385

#140

#141

#1414

#1415

#142

#1420

#143

#1439

#1439

#145

#146

#1467

#1471

#1473

#1476

#1479

#148

#1483

#1493

#1494

#1498

#150

#1500

#1502

#1520

#1522

#1525

#1527

#1532

#1538

#154

#1545

#1547

#155

#1552

#1553

#1553

#156

#1562

#1564

#1565

#1575

#1578

#1578

#1581

#1582

#1583

#1586

#1590

#1590

#1592

#1593

#16

#1606

#1612

#1614

#1617

#1617

#1621

#1622

#1624

#1627

#1628

#1628

#163

#1630

#1633

#1641

#1643

#1643

#1645

#1647

#1647

#1649

#1654

#1654

#1655

#1659

#1659

#1662

#1673

#1674

#1677

#1679

#1679

#1680

#1683

#1685

#1687

#1688

#1688

#175

#179

#183

#19

#191

#195

#199

#2

#208

#212

#212

#213

#213

#214

#216

#218

#227

#228

#23

#233

#237

#24

#240

#240

#244

#250

#255

#256

#263

#268

#282

#284

#286

#29

#290

#290

#297

#30

#300

#302

#31

#315

#32

#322

#323

#33

#334

#335

#35

#352

#367

#382

#389

#401

#413

#418

#419

#421

#425

#427

#429

#434

#437

#440

#447

#450

#459

#475

#489

#494

#497

#501

#502

#510

#52

#525

#528

#53

#533

#537

#542

#548

#550

#554

#555

#556

#561

#571

#580

#581

#595

#6

#603

#603

#609

#612

#613

#62

#625

#626

#636

#637

#645

#647

#65

#651

#652

#653

#659

#659

#66

#660

#668

#669

#679

#681

#685

#693

#7

#702

#714

#729

#731

#738

#74

#746

#747

#750

#768

#768

#771

#772

#800

#801

#804

#804

#805

#808

#812

#813

#816

#817

#82

#821

#828

#83

#841

#842

#853

#853

#858

#87

#874

#875

#876

#88

#880

#882

#891

#898

#910

#913

#914

#915

#933

#94

#95

#964

#967

#968

#986

#989

#99

#993

v0.1-attn-weights

v0.1-cadene

v0.1-coat-weights

v0.1-dnf-weights

v0.1-effv2-weights

v0.1-ger-weights

v0.1-hrnet

v0.1-mvit-weights

v0.1-pit-weights

v0.1-regnet

v0.1-repvgg-weights

v0.1-res2net

v0.1-resnest

v0.1-rexnet

v0.1-rs-weights

v0.1-rsb-weights

v0.1-selecsls

v0.1-tpu-weights

v0.1-tresnet

v0.1-vitjx

v0.1-vt3p-weights

v0.1-weights

v0.1-weights-maxx

v0.1-weights-morevit

v0.1-weights-swinv2

v0.1.1

v0.2.1

v0.3.3

v0.3.4

v0.4.12

v0.4.5

v0.4.9

v0.5.4

v0.6.11

v0.6.12

v0.6.5

v0.6.7

v0.8.10dev0

v0.8.13dev0

v0.8.2dev0

v0.8.6dev0

Mono Color

• c11973602d
  Merge pull request #19 from rwightman/resnet-refactor Ross Wightman 2019-07-20 08:47:21 -0700
• d97510fd4a Add new results csv and update README with 3 new ResNet weight results #19 Ross Wightman 2019-07-20 08:45:05 -0700
• 3d9be78fc6 A bit more ResNet cleanup. * add inplace=True back * minor comment improvements * few clarity changes Ross Wightman 2019-07-19 16:44:35 -0700
• 33436fafad Add weights for ResNeXt50d model Ross Wightman 2019-07-19 14:09:10 -0700
• e78cd79073 Move ResNet additions for Gluon into main ResNet impl. Add ResNet-26 and ResNet-26d models with weights. Ross Wightman 2019-07-14 18:17:35 -0700
• c37cb0f2b7 ImageNet-1k vs ImageNet-v2 comparison Ross Wightman 2019-07-06 17:19:56 -0700
• 6cdf35e670 Add explicit half/fp16 support to loader and validation script Ross Wightman 2019-07-05 13:50:12 -0700
• 5684c6af32
  Update README.md Ross Wightman 2019-07-04 10:52:03 -0700
• a6b2f6eca5 Update README, bump version Ross Wightman 2019-07-03 22:43:22 -0700
• 2a1d09b4c0 Add result csv with summary of all pretrained models Ross Wightman 2019-07-03 22:34:36 -0700
• edb425ea48 Add crop_pct arg to validate, extra fields to csv output, 'all' filters pretrained Ross Wightman 2019-07-03 22:28:07 -0700
• 949b7a81c4 Fix typo in Densenet default resolutions Ross Wightman 2019-07-03 22:11:26 -0700
• da52fcf78a Add NASNet-Large model Ross Wightman 2019-07-03 22:10:50 -0700
• f632190fd9 Add .gitattributes Ross Wightman 2019-07-01 21:44:50 -0700
• 5bc0ed3d7c Grammar Ross Wightman 2019-07-01 15:15:21 -0700
• c89acdfc02
  Update EffResNetComparison.ipynb Ross Wightman 2019-07-01 14:57:11 -0700
• 73642f7462 Adjust notebook path Ross Wightman 2019-07-01 14:16:18 -0700
• c30f625215 Update EffResNetComparison Ross Wightman 2019-07-01 14:07:47 -0700
• c4916897f8 EfficientNet vs ResNet comparison in Colaboratory Ross Wightman 2019-07-01 13:55:47 -0700
• 6057496409 Register dpn107 Ross Wightman 2019-06-30 09:57:06 -0700
• 3d1a66b6fc Version 0.1.6 Ross Wightman 2019-06-30 09:55:23 -0700
• a6878b5218 Fix DPN config keys that I broke Ross Wightman 2019-06-30 09:54:52 -0700
• 8a05b8d555
  Merge pull request #16 from rwightman/misc-epoch Ross Wightman 2019-06-29 16:59:11 -0700
• 9b0070edc9 Add two comments back, fix typo #16 Ross Wightman 2019-06-29 16:44:25 -0700
• 188aeae8f4 Bump version 0.1.4 Ross Wightman 2019-06-29 16:17:54 -0700
• c3287aafb3 Slight improvement in EfficientNet-B2 native PyTorch weights Ross Wightman 2019-06-29 16:17:06 -0700
• b8762cc67d Model updates. Add my best ResNet50 weights top-1=78.47. Add some other torchvision weights. * Remove some models that don't exist as pretrained an likely never will (se)resnext152 * Add some torchvision weights as tv_ for models that I have added better weights for * Add wide resnet recently added to torchvision along with resnext101-32x8d * Add functionality to model registry to allow filtering on pretrained weight presence Ross Wightman 2019-06-29 15:37:42 -0700
• 65a634626f Switch random erasing to doing normal_() on CPU to avoid instability, remove a debug print Ross Wightman 2019-06-29 10:03:13 -0700
• c6b32cbe73 A number of tweaks to arguments, epoch handling, config * reorganize train args * allow resolve_data_config to be used with dict args, not just arparse * stop incrementing epoch before save, more consistent naming vs csv, etc * update resume and start epoch handling to match above * stop auto-incrementing epoch in scheduler Ross Wightman 2019-06-28 13:49:20 -0700
• 9d653b68a2 Make drop_connect rate scaling match official impl. Fixes #14 Ross Wightman 2019-06-25 09:30:36 -0700
• 13c19e213d Add native PyTorch EfficientNet B1 and B2 weights. Not quite where I want them, but hitting the brick wall and moving on to other projects... Ross Wightman 2019-06-24 13:12:04 -0700
• a0275cfa2f Fix arg positions in two entrypoint aliases Ross Wightman 2019-06-24 08:25:14 -0700
• fe59249701 Bump version to 0.1.2 Ross Wightman 2019-06-23 18:27:30 -0700
• 171c0b88b6 Add model registry and model listing fns, refactor model_factory/create_model fn Ross Wightman 2019-06-23 18:22:16 -0700
• 8512436436 Add instagram pretrained ResNeXt models from https://pytorch.org/hub/facebookresearch_WSL-Images_resnext/, update README Ross Wightman 2019-06-23 12:29:02 -0700
• 1202d053bc markdown in setup.py v0.1.1 Ross Wightman 2019-06-21 14:08:43 -0700
• 87b92c528e Some pretrianed URL changes * host some of Cadene's weights on github instead of .fr for speed * add my old port of ensemble adversarial inception resnet v2 * switch to my TF port of normal inception res v2 and change FC layer back to 'classif' for compat with ens_adv Ross Wightman 2019-06-21 13:57:08 -0700
• 827a3d6010 Add current checkpoints output back to CheckpointSaver (via logger) v0.1-cadene Ross Wightman 2019-06-21 11:57:43 -0700
• 63961b36a2 Missed pnasnet entrypoint Ross Wightman 2019-06-20 23:34:20 -0700
• 6cc214bd7a Consistency in model entrypoints * move pretrained entrypoint arg to first pos to be closer to torchvision/hub * change DPN weight URLS to my github location Ross Wightman 2019-06-20 23:29:44 -0700
• b20bb58284 Distributed tweaks * Support PyTorch native DDP as fallback if APEX not present * Support SyncBN for both APEX and Torch native (if torch >= 1.1) * EMA model does not appear to need DDP wrapper, no gradients, updated from wrapped model Ross Wightman 2019-06-20 22:27:09 -0700
• 6fc886acaf Remove all prints, change most to logging calls, tweak alignment of batch logs, improve setup.py Ross Wightman 2019-06-20 17:29:25 -0700
• ebafb4238e
  Replace `getattr` with `hasattr` in `get_state_dict` #13 Maxim Berman 2019-06-20 16:12:37 -0700
• 1d7f2d93a6 Python >= 3.6 requirement Ross Wightman 2019-06-19 17:28:08 -0700
• aa4354f466 Big re-org, working towards making pip/module as 'timm' Ross Wightman 2019-06-19 17:19:37 -0700
• 871f4c1b0c Change some absolute imports to relative Ross Wightman 2019-06-19 16:08:58 -0700
• b9f8d40b10 Fix pretrained override logic for validate, checkpoint always trump pretrained flag during model create Ross Wightman 2019-06-19 15:23:16 -0700
• 0e1fd11ad8
  Merge pull request #12 from rwightman/ema-cleanup Ross Wightman 2019-06-18 12:37:16 -0700
• 1019414fd2 Add ported Tensorflow EfficientNet B4/B5 weights Ross Wightman 2019-06-18 09:50:36 -0700
• c9a61b7d98 Add gitignore Ross Wightman 2019-06-18 08:41:29 -0700
• c6b8259ad6
  Update README.md Ross Wightman 2019-06-14 10:51:54 -0700
• f3134973b5 Add the MnasNet-B1 variant weights, add A1/B1 model names as in the stand-alone repo, remove a bit of unused code Ross Wightman 2019-06-14 09:08:02 -0700
• c1a84ecb22 dataset not passed through PrefetchLoader for inference script. Fix #10 * also, make top5 configurable for lower class count cases Ross Wightman 2019-06-11 21:59:56 -0700
• 2060e433c0 Add native PyTorch weights for SE-MnasNet aka MnasNet-A1 Ross Wightman 2019-06-11 08:54:54 -0700
• 7dab6d1ec7 Default to img_size in model default_cfg, defer output folder creation until later in the init sequence #12 Ross Wightman 2019-06-10 13:34:42 -0700
• 7d17394bdc Add native PyTorch weights for EfficientNet-B0 w/ top-1 > 76.9 * also add pooling details to default cfg for efficiennets so testtimepool wrapper works Ross Wightman 2019-06-10 13:11:36 -0700
• 9bcd65181b Add exponential moving average for model weights + few other additions and cleanup * ModelEma class added to track an EMA set of weights for the model being trained * EMA handling added to train, validation and clean_checkpoint scripts * Add multi checkpoint or multi-model validation support to validate.py * Add syncbn option (APEX) to train script for experimentation * Cleanup interface of CheckpointSaver while adding ema functionality Ross Wightman 2019-06-07 15:39:36 -0700
• 105d5702d7
  Update README.md Ross Wightman 2019-06-04 10:23:36 -0700
• 6688d0669f PyTorch trained MobileNetV3 weights that match/best paper. Update fbnetc weights to improved copy. Ross Wightman 2019-06-04 09:51:05 -0700
• 89e77ec049
  Merge pull request #7 from zhunzhong07/patch-2 Ross Wightman 2019-06-03 13:00:19 -0700
• b32361c9f8
  Correct name #7 Zhun Zhong 2019-06-03 20:33:04 +1000
• 0dab9cdedb
  Avoid shadowing builtin: input --> curr_input #6 cclauss 2019-06-02 20:08:10 +0200
• ff99625603 Missed a few models in the model/_all_ list for senet Ross Wightman 2019-05-31 10:51:18 -0700
• 99122aac1c Replace ResNet-34 default weights with a great result from my experiments. Ross Wightman 2019-05-31 10:34:37 -0700
• 48ab3cf070 Add Apache LICENSE file Ross Wightman 2019-05-31 07:24:03 -0700
• 4bb5e9b224 Ported Tensorflow pretrained EfficientNet weights and some model cleanup * B0-B3 weights ported from TF with close to paper accuracy * Renamed gen_mobilenet to gen_efficientnet since scaling params go well beyond 'mobile' specific * Add Tensorflow preprocessing option for closer images to source repo Ross Wightman 2019-05-30 17:52:52 -0700
• 4efecfdc47 Add drop_connect impl to try during training, fix a few comments v0.1-weights Ross Wightman 2019-05-30 08:42:12 -0700
• 0fc4cca2ff Cleanup unused member with old name Ross Wightman 2019-05-29 23:44:01 -0700
• 7a6d61566e Add EfficientNet impl, change existing depth_multipler -> channel_multiplier as definitions have been confused Ross Wightman 2019-05-29 23:32:52 -0700
• 6bff9c75dc Cleanup model_factory imports, consistent __all__ for models, fixed inception_v4 weight url Ross Wightman 2019-05-28 21:39:54 -0700
• e6c14427c0 More appropriate/correct loss name Ross Wightman 2019-05-28 16:56:07 -0700
• 99ab1b1276
  Update README.md Ross Wightman 2019-05-26 18:48:35 -0700
• c2de3a922b
  Update README.md Ross Wightman 2019-05-26 18:46:12 -0700
• 337702d51b
  Merge pull request #2 from zhunzhong07/patch-1 Ross Wightman 2019-05-26 09:31:40 -0700
• 127487369f
  Fix bug for prefetcher #2 Zhun Zhong 2019-05-26 15:37:19 +1000
• e8cf619005
  Update gluon_resnet.py Ross Wightman 2019-05-24 00:01:37 -0700
• 7419e9835f Add MxNet Gluon ResNet variants w/ converted pretrained weights. Very well trained set of models. Ross Wightman 2019-05-23 23:52:23 -0700
• 2da0b4dbc1 Add inception_v3 models via torchvision, 4 different pretrained weight choices Ross Wightman 2019-05-23 10:59:31 -0700
• af1a68d2e1
  Update README.md Ross Wightman 2019-05-20 14:55:56 -0700
• 8ceceef889 Densenet should default to bicubic interpolation, update model links to 'cpu fix' from long ago Ross Wightman 2019-05-20 11:45:29 -0700
• a4516fe0fb Add pretrained weights for FBNet-C Ross Wightman 2019-05-20 11:05:25 -0700
• a72e9b67ee
  Update README.md Ross Wightman 2019-05-18 22:27:42 -0700
• 4d2056722a Mixup and prefetcher improvements * Do mixup in custom collate fn if prefetcher enabled, reduces performance impact * Move mixup code to own file * Add arg to disable prefetcher * Fix no cuda transfer when prefetcher off * Random erasing when prefetcher off wasn't changed to match new args, fixed * Default random erasing to off (prob = 0.) for train Ross Wightman 2019-05-18 22:17:51 -0700
• 780c0a96a4 Change args for RandomErasing so only one required for pixel/color mode Ross Wightman 2019-05-18 12:29:30 -0700
• 76539d905e Some transform/data/loader refactoring, hopefully didn't break things * factor out data related constants to own file * move data related config helpers to own file * add a variant of RandomResizeCrop that randomizes interpolation method * remove old Numpy version of RandomErasing * cleanup torch version of RandomErasing and use it in either GPU loader batch mode or single image cpu Transform Ross Wightman 2019-05-16 22:52:17 -0700
• e3377b0409 Add some info to README Ross Wightman 2019-05-15 09:54:07 -0700
• db8ad25a23 MobileNetV3 appears correct based on paper update, cleaned up comments and compacted last block def Ross Wightman 2019-05-15 08:53:27 -0700
• 20d66beead Move RMSpropTF another step closer to Tensorflow impl * init square_avg with one instead of zero as per TF * match TF order of ops for square_avg accumulation * move LR scaling to momentum buffer accumulator as per TF * add decoupled weight decay flag (not in TF) Ross Wightman 2019-05-14 18:30:46 -0700
• 89147a91e6 Remove adabound optimizer, never got it working well on larger datasets Ross Wightman 2019-05-14 18:29:34 -0700
• 0a84dd5890
  Merge pull request #1 from cclauss/patch-1 Ross Wightman 2019-05-14 08:52:37 -0700
• 51f85702a7
  Identity is not the same thing as equality in Python #1 cclauss 2019-05-14 17:19:57 +0200
• fee607edf6 Mixup implemention in progress * initial impl w/ label smoothing converging, but needs more testing Ross Wightman 2019-05-13 19:05:40 -0700
• c3fbdd4655 Fix efficient head for MobileNetV3 Ross Wightman 2019-05-11 11:23:11 -0700
• 17da1adaca A few MobileNetV3 tweaks * fix expansion ratio on early block * change comment re stride mistake in paper * fix rounding not being called properly for all multipliers != 1.0 Ross Wightman 2019-05-11 10:23:40 -0700
• 6523e4abe4 More appropriate name for se channel Ross Wightman 2019-05-10 23:32:18 -0700
• db056d97e2 Add MobileNetV3 and associated changes hard-swish, hard-sigmoid, efficient head, etc Ross Wightman 2019-05-10 23:28:13 -0700
• 02abeb95bf Add the url for tflite mnasnet ported weights Ross Wightman 2019-04-28 23:21:35 -0700
• 0956cd4b66
  Update README.md Ross Wightman 2019-04-28 18:06:20 -0700
• 4663fc2132 Add support for tflite mnasnet pretrained weights and included spnasnet pretrained weights of my own. * tensorflow 'SAME' padding support added to GenMobileNet models for tflite pretrained weights * folded batch norm support (made batch norm optional and enable conv bias) for tflite pretrained weights * add url for spnasnet1_00 weights that I recently trained * fix SE reduction size for semnasnet models Ross Wightman 2019-04-28 17:45:07 -0700
• afb357ff68 Make genmobilenet weight init switchable, fix fan_out in google style linear init Ross Wightman 2019-04-22 17:46:17 -0700