Migrations
/
pytorch-image-models
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Commit Graph

  • f2e14685a8 Add force-cpu flag for train/validate, fix CPU fallback for device init, remove old force cpu flag for EMA model weights Ross Wightman 2021-08-22 13:15:27 -0700
  • a8b65695f1 Add resnet26ts and resnext26ts models for non-attn baselines Ross Wightman 2021-08-21 12:42:10 -0700
  • a5a542f17d Fix typo Ross Wightman 2021-08-20 17:47:23 -0700
  • 925e102982 Update attention / self-attn based models from a series of experiments: * remove dud attention, involution + my swin attention adaptation don't seem worth keeping * add or update several new 26/50 layer ResNe(X)t variants that were used in experiments * remove models associated with dead-end or uninteresting experiment results * weights coming soon... Ross Wightman 2021-08-20 16:13:11 -0700
  • acd6c687fd git push origin masterMerge branch 'yohann84L-fix_accuracy' Ross Wightman 2021-08-19 14:26:23 -0700
  • d667351eac Tweak accuracy topk safety. Fix #807 Ross Wightman 2021-08-19 14:18:53 -0700
  • 2ee398d501 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-08-19 13:59:53 -0700
  • 35c9740826 Fix accuracy when topk > num_classes #817 Yohann Lereclus 2021-08-19 11:58:59 +0200
  • d54cdedc86 add two last outputs to feature_info #816 Yohann Lereclus 2021-08-19 10:56:53 +0200
  • a16a753852 Add lamb/lars to optim init imports, remove stray comment Ross Wightman 2021-08-18 22:55:02 -0700
  • c207e02782 MOAR optimizer changes. Woo! Ross Wightman 2021-08-18 22:20:35 -0700
  • 42c1f0cf6c Fix lars tests Ross Wightman 2021-08-18 21:05:34 -0700
  • f4fb068b11 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-08-18 17:25:00 -0700
  • a426511c95 More optimizer cleanup. Change all to no longer use .data. Improve (b)float16 use with adabelief. Add XLA compatible Lars. Ross Wightman 2021-08-18 17:21:56 -0700
  • b0265ef8a6 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-08-18 11:21:21 -0700
  • 9541f4963b One more scalar -> tensor fix for lamb optimizer Ross Wightman 2021-08-18 11:20:25 -0700
  • 0d82876132 Add comment for reference re PyTorch XLA 'race' issue Ross Wightman 2021-08-18 11:15:32 -0700
  • b76b48e8e9 Update optimizer creation for master optimizer changes Ross Wightman 2021-08-18 11:14:51 -0700
  • f98662b9c9 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-08-18 09:31:10 -0700
  • 8f68193c91
         Update lamp.py comment Ross Wightman 2021-08-18 09:27:40 -0700
  • 4d284017b8
         Merge pull request #813 from rwightman/opt_cleanup Ross Wightman 2021-08-18 09:12:00 -0700
  • a6af48be64 add madgradw optimizer #813 Ross Wightman 2021-08-17 22:19:27 -0700
  • 55fb5eedf6 Remove experiment from lamb impl Ross Wightman 2021-08-17 21:48:26 -0700
  • 8a9eca5157 A few optimizer comments, dead import, missing import Ross Wightman 2021-08-17 18:01:33 -0700
  • 959eaff121 Add optimizer tests and update testing to pytorch 1.9 Ross Wightman 2021-08-17 17:59:15 -0700
  • ac469b50da Optimizer improvements, additions, cleanup * Add MADGRAD code * Fix Lamb (non-fused variant) to work w/ PyTorch XLA * Tweak optimizer factory args (lr/learning_rate and opt/optimizer_name), may break compat * Use newer fn signatures for all add,addcdiv, addcmul in optimizers * Use upcoming PyTorch native Nadam if it's available * Cleanup lookahead opt * Add optimizer tests * Remove novograd.py impl as it was messy, keep nvnovograd * Make AdamP/SGDP work in channels_last layout * Add rectified adablief mode (radabelief) * Support a few more PyTorch optim, adamax, adagrad Ross Wightman 2021-08-17 17:51:20 -0700
  • 93b27d69e7 Fix RuntimeError: Integer division of tensors using div or / is no longer supported, and in a future release div will perform true division as in Python 3. Use true_divide or floor_divide (// in Python) instead. #812 danil328 2021-08-17 14:19:59 +0300
  • b1d570b0ad Fix RuntimeError: Integer division of tensors using div or / is no longer supported, and in a future release div will perform true division as in Python 3. Use true_divide or floor_divide (// in Python) instead. danil328 2021-08-17 13:57:44 +0300
  • 368211d19a
         Merge pull request #805 from Separius/patch-1 Ross Wightman 2021-08-15 12:51:43 -0700
  • 87e2033b1d topk bug fix for binary classification #808 Reddy, Yeshwanth 2021-08-15 17:43:22 +0530
  • abf3e044bb
         Update scheduler_factory.py #805 Sepehr Sameni 2021-08-14 22:53:17 +0200
  • cb621e0f00 Remove print, arg order Ross Wightman 2021-08-13 12:54:12 -0700
  • b974d85026 Merge branch 'bits_and_tpu' of github.com:rwightman/pytorch-image-models into bits_and_tpu Ross Wightman 2021-08-13 12:49:16 -0700
  • c06c739901 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-08-13 12:49:04 -0700
  • 40457e5691 Transforms, augmentation work for bits, add RandomErasing support for XLA (pushing into transforms), revamp of transform/preproc config, etc ongoing... Ross Wightman 2021-08-13 12:45:43 -0700
  • 3cdaf5ed56 Add `mmax` config key to auto_augment for increasing upper bound of RandAugment magnitude beyond 10. Make AugMix uniform sampling default not override config setting. Ross Wightman 2021-08-12 15:39:05 -0700
  • 09a45ab592
         fix a typo in ### Select specific feature levels or limit the stride #801 Yonghye Kwon 2021-08-13 01:36:48 +0900
  • 1042b8a146 Add non fused LAMB optimizer option Ross Wightman 2021-08-09 13:13:43 -0700
  • 01cb46a9a5 Add gc_efficientnetv2_rw_t weights (global context instead of SE attn). Add TF XL weights even though the fine-tuned ones don't validate that well. Change default arg for GlobalContext to use scal (mul) mode. Ross Wightman 2021-08-07 16:45:29 -0700
  • 7fd6273104
         Merge 55b135d5c6 into bd56946676 #768 Ross Wightman 2021-07-28 13:31:17 -0400
  • bd56946676
         Update README.md Ross Wightman 2021-07-28 09:00:48 -0700
  • abf02db3d7 [M] making PatchEmbed safer #772 Carl-Johann SIMON-GABRIEL 2021-07-25 12:02:31 +0200
  • 7be92e30aa Removed extraneous activation in SEResNetBlock #771 ayhyap 2021-07-24 23:38:52 +0800
  • 55b135d5c6 Add layer scale + affine option to perceiver #768 perceiver Ross Wightman 2021-07-22 14:27:27 -0700
  • d3f7440650 Add EfficientNetV2 XL model defs Ross Wightman 2021-07-22 13:15:24 -0700
  • 820c262f33 Don't exclude perceivers from tests (yet) Ross Wightman 2021-07-21 13:13:53 -0700
  • 77698d80a5 Initial Perceiver impl. WIP Ross Wightman 2021-07-21 13:12:11 -0700
  • ef1e2e12be Attempt to fix xcit test failures on github runner by filter largest models Ross Wightman 2021-07-13 16:33:04 -0700
  • 72b227dcf5
         Merge pull request #750 from drjinying/master Ross Wightman 2021-07-13 11:01:20 -0700
  • 2907c1f967
         Merge pull request #746 from samarth4149/master Ross Wightman 2021-07-13 10:55:54 -0700
  • 5aca7c01e5 Update README.md Ross Wightman 2021-07-12 13:33:02 -0700
  • 763329f23f Merge branch 'alexander-soare-xcit' Ross Wightman 2021-07-12 13:28:15 -0700
  • 748ab852ca Allow act_layer switch for xcit, fix in_chans for some variants Ross Wightman 2021-07-12 13:27:29 -0700
  • 20b2d4b69d Use bicubic interpolation in resize_pos_embed() #750 Ying Jin 2021-07-11 22:08:07 -0700
  • d3255adf8e Merge branch 'xcit' of https://github.com/alexander-soare/pytorch-image-models into alexander-soare-xcit Ross Wightman 2021-07-12 08:30:30 -0700
  • f8039c7492 Fix gc effv2 model cfg name Ross Wightman 2021-07-11 12:14:31 -0700
  • 3a55a30ed1 add notes from author #747 Alexander Soare 2021-07-11 14:25:58 +0100
  • 899cf84ccc bug fix - missing _dist postfix for many of the 224_dist models Alexander Soare 2021-07-11 12:41:51 +0100
  • 623e8b8eb8 wip xcit Alexander Soare 2021-07-08 21:05:44 +0100
  • 392368e210 Add efficientnetv2_rw_t defs w/ weights, and gc variant, as well as gcresnet26ts for experiments. Version 0.4.13 Ross Wightman 2021-07-09 16:46:52 -0700
  • daab57a6d9 1. Added a simple multi step LR scheduler #746 samarth 2021-07-09 16:18:27 -0400
  • 6d8272e92c Add SAM pretrained model defs/weights for ViT B16 and B32 models. Ross Wightman 2021-07-08 11:23:55 -0700
  • ad04c561f1 Updated interpolation methods torchvision InterpolationMode #738 Maxwell Aladago 2021-07-07 08:25:30 -0400
  • ee4d8fc69a Remove unecessary line from nest post refactor Ross Wightman 2021-07-05 21:22:46 -0700
  • c8ec1ffcb9 Merge branch 'alexander-soare-nested_transformer' Ross Wightman 2021-07-05 18:22:50 -0700
  • 8165cacd82 Realized LayerNorm2d won't work in all cases as is, fixed. Ross Wightman 2021-07-05 18:21:34 -0700
  • 81cd6863c8 Move aggregation (convpool) for nest into NestLevel, cleanup and enable features_only use. Finalize weight url. Ross Wightman 2021-07-05 18:20:49 -0700
  • 6ae0ac6420 Merge branch 'nested_transformer' of https://github.com/alexander-soare/pytorch-image-models into alexander-soare-nested_transformer Ross Wightman 2021-07-03 12:45:26 -0700
  • 7b8a0017f1 wip to review #731 Alexander Soare 2021-07-03 12:10:12 +0100
  • b11d949a06 wip checkpoint with some feature extraction work Alexander Soare 2021-07-03 11:45:19 +0100
  • 23bb72ce5e nested_transformer wip Alexander Soare 2021-06-24 21:02:13 +0100
  • 7919053425
         Merge pull request #729 from bryant1410/patch-3 Ross Wightman 2021-07-02 10:25:40 -0700
  • 49b38a51e3
         Add color highlighting to BibTeX entry in README #729 Santiago Castro 2021-06-30 14:07:16 -0700
  • 7096b52a61 Remove sotabench files, no longer working / maintained v0.4.12 Ross Wightman 2021-06-30 09:25:58 -0700
  • d10b071a28 Update results csvs w/ latest ViT, ResMLP, and NfNet-L2 weights present Ross Wightman 2021-06-30 09:10:35 -0700
  • 766b4d3262 Fix features for resnetv2_50t Ross Wightman 2021-06-28 15:56:24 -0700
  • e8045e712f Fix BatchNorm for ResNetV2 non GN models, add more ResNetV2 model defs for future experimentation, fix zero_init of last residual for pre-act. Ross Wightman 2021-06-28 10:52:45 -0700
  • 02aaa785b9
         Update README.md Ross Wightman 2021-06-23 14:08:16 -0700
  • 7606bdf9e8
         Merge pull request #714 from rwightman/vit_and_bit_test_fixes Ross Wightman 2021-06-23 11:36:55 -0700
  • 5e95ced5a7 timm bits checkpoint support for avg_checkpoints.py Ross Wightman 2021-06-23 11:10:05 -0700
  • 56ed0a0b63 Merge branch 'vit_and_bit_test_fixes' into bits_and_tpu Ross Wightman 2021-06-23 11:02:21 -0700
  • 20a2be14c3 Add gMLP-S weights, 79.6 top-1 #714 vit_and_bit_test_fixes Ross Wightman 2021-06-23 10:40:30 -0700
  • 85f894e03d Fix ViT in21k representation (pre_logits) layer handling across old and new npz checkpoints Ross Wightman 2021-06-23 10:38:34 -0700
  • b41cffaa93 Fix a few issues loading pretrained vit/bit npz weights w/ num_classes=0 __init__ arg. Missed a few other small classifier handling detail on Mlp, GhostNet, Levit. Should fix #713 Ross Wightman 2021-06-22 23:16:05 -0700
  • dc422820ec
         Update README.md Ross Wightman 2021-06-20 22:12:56 -0700
  • 79927baaec
         Merge pull request #702 from rwightman/cleanup_xla_model_fixes Ross Wightman 2021-06-20 17:49:14 -0700
  • 9c9755a808 AugReg release #702 cleanup_xla_model_fixes Ross Wightman 2021-06-20 17:46:06 -0700
  • 381b279785 Add hybrid model fwds back Ross Wightman 2021-06-19 22:28:44 -0700
  • 26f04a8e3e Fix a weight link Ross Wightman 2021-06-19 16:39:36 -0700
  • 8f4a0222ed Add GMixer-24 MLP model weights, trained w/ TPU + PyTorch XLA Ross Wightman 2021-06-18 16:49:28 -0700
  • 4c09a2f169 Bump version 0.4.12 Ross Wightman 2021-06-18 16:17:34 -0700
  • b319eb5b5d Update ViT weights, more details to be added before merge. Ross Wightman 2021-06-18 16:16:49 -0700
  • 8257b86550 Fix up resnetv2 bit/bitm model default res Ross Wightman 2021-06-18 16:16:06 -0700
  • 1228f5a3d8 Add BiT distilled 50x1 and teacher 152x2 models from 'A good teacher is patient and consistent' paper. Ross Wightman 2021-06-18 11:40:33 -0700
  • 511a8e8c96 Add official ResMLP weights. Ross Wightman 2021-06-14 17:01:12 -0700
  • b9cfb64412 Support npz custom load for vision transformer hybrid models. Add posembed rescale for npz load. Ross Wightman 2021-06-14 12:31:44 -0700
  • 8319e0c373 Add file docstring to std_conv.py Ross Wightman 2021-06-13 12:31:06 -0700
  • 0020268d9b Try lower max size for non_std default_cfg test Ross Wightman 2021-06-12 23:31:24 -0700
  • 4d96165989 Merge branch 'master' into cleanup_xla_model_fixes Ross Wightman 2021-06-12 23:19:25 -0700
  • 8880f696b6 Refactoring, cleanup, improved test coverage. * Add eca_nfnet_l2 weights, 84.7 @ 384x384 * All 'non-std' (ie transformer / mlp) models have classifier / default_cfg test added * Fix #694 reset_classifer / num_features / forward_features / num_classes=0 consistency for transformer / mlp models * Add direct loading of npz to vision transformer (pure transformer so far, hybrid to come) * Rename vit_deit* to deit_* * Remove some deprecated vit hybrid model defs * Clean up classifier flatten for conv classifiers and unusual cases (mobilenetv3/ghostnet) * Remove explicit model fns for levit conv, just pass in arg Ross Wightman 2021-06-12 16:40:02 -0700