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

Commit Graph

  • be53107e8a Update README, ensure vit excluded from all tests (not ready) #250 Ross Wightman 2020-10-13 14:51:08 -0700
  • f31933cb37 Initial Vision Transformer impl w/ patch and hybrid variants. Refactor tuple helpers. Ross Wightman 2020-10-13 13:33:44 -0700
  • 9305313291 Default to old checkpoint format for now, still want compatibility with older torch ver for released models Ross Wightman 2020-10-13 12:58:04 -0700
  • a4d8fea61e Add model based wd skip support. Improve cross version compat of optimizer factory. Fix #247 Ross Wightman 2020-10-13 12:49:47 -0700
  • 80078c47bb Add Adafactor and Adahessian optimizers, cleanup optimizer arg passing, add gradient clipping support. Ross Wightman 2020-10-09 17:24:08 -0700
  • fcb6258877 Add missing leaky_relu layer factory defn, update Apex/Native loss scaler interfaces to support unscaled grad clipping. Bump ver to 0.2.2 for pending release. Ross Wightman 2020-10-02 16:17:42 -0700
  • 186075ef03
         Merge pull request #244 from hollance/master Ross Wightman 2020-10-02 15:53:48 -0700
  • f04bdc8c8e don't forget this file #244 Matthijs Hollemans 2020-10-02 23:23:44 +0200
  • 8ffdc5910a test_time_pool would be set to a non-False value even if test-time pooling is not available Matthijs Hollemans 2020-10-02 22:56:15 +0200
  • 4be5b51e0a Missed moving some seresnet -> legacy in sotabench. Check sotabench cache. Ross Wightman 2020-09-25 15:51:04 -0700
  • e8e2d9cabf Add DropPath (stochastic depth) to ReXNet and VoVNet. RegNet DropPath impl tweak and dedupe se args. Ross Wightman 2020-09-24 18:20:36 -0700
  • e8ca45854c More models in sotabench, more control over sotabench run, dataset filename extraction consistency Ross Wightman 2020-09-24 15:55:28 -0700
  • 9c406532bd Add EfficientNet-EdgeTPU-M (efficientnet_em) model trained natively in PyTorch. More sotabench fiddling. Ross Wightman 2020-09-23 17:12:07 -0700
  • d1b2136669 remove unecessary imports datamining108 2020-09-21 20:55:21 +0900
  • 08a6be1390 update post-quantization feature with various model supports datamining108 2020-09-21 20:41:18 +0900
  • 3681c5c4dd Another sotabench.py debug iter Ross Wightman 2020-09-18 23:46:46 -0700
  • 08029852d9 Sotabench debugging Ross Wightman 2020-09-18 22:14:51 -0700
  • c40384f5bd Add ResNet weights. 80.5 (top-1) ResNet-50-D, 77.1 ResNet-34-D, 72.7 ResNet-18-D. Ross Wightman 2020-09-18 12:05:37 -0700
  • e39bf6ef59
         Merge pull request #237 from rwightman/utils_cleanup Ross Wightman 2020-09-11 13:29:18 -0700
  • 47a7b3b5b1 More flexible mixup mode, add 'half' mode. #237 Ross Wightman 2020-09-07 20:03:06 -0700
  • 532e3b417d Reorg of utils into separate modules Ross Wightman 2020-09-07 13:56:00 -0700
  • 9ce42d5c5a
         Update README.md Ross Wightman 2020-09-03 10:48:25 -0700
  • 0729dbe865
         Update README.md Ross Wightman 2020-09-03 10:48:11 -0700
  • 33f8a1bf36 Updated README, add wide_resnet50_2 and seresnext50_32x4d weights Ross Wightman 2020-09-03 10:45:17 -0700
  • 5247eb37a7
         Merge pull request #233 from rwightman/torchamp Ross Wightman 2020-09-02 17:42:41 -0700
  • 751b0bba98 Add global_pool (--gp) arg changes to allow passing 'fast' easily for train/validate to avoid channels_last issue with AdaptiveAvgPool #233 Ross Wightman 2020-09-02 16:13:47 -0700
  • 9c297ec67d Cleanup Apex vs native AMP scaler state save/load. Cleanup CheckpointSaver a bit. Ross Wightman 2020-09-02 15:12:59 -0700
  • 80c9d9cc72 Add 'fast' global pool option, remove redundant SEModule from tresnet, normal one is now 'fast' Ross Wightman 2020-09-02 09:11:48 -0700
  • 90a01f47d1 hrnet features_only pretrained weight loading issue. Fix #232. Ross Wightman 2020-09-01 17:37:55 -0700
  • 110a7c4982 AdaptiveAvgPool2d -> mean((2,3)) for all SE/attn layers to avoid NaN with AMP + channels_last layout. See https://github.com/pytorch/pytorch/issues/43992 Ross Wightman 2020-09-01 16:01:36 -0700
  • c2cd1a332e Improve torch amp support and add channels_last support for train/validate scripts Ross Wightman 2020-08-31 17:58:16 -0700
  • 1d34a0a851 Merge branch 'master' of https://github.com/tgisaturday/pytorch-image-models into torchamp Ross Wightman 2020-08-30 18:16:47 -0700
  • 6d158adfe9
         Update README.md Ross Wightman 2020-08-26 15:15:51 -0700
  • 5f563ca4df fix save_checkpoint bug with native amp #228 datamining99 2020-08-22 11:31:50 +0900
  • d98967ed5d add support for native torch AMP in torch 1.6 #227 datamining99 2020-08-22 09:44:23 +0900
  • 470220b1f4 Fix MobileNetV3 crash with global_pool='', output consistent with other models but not equivalent due to efficient head. Ross Wightman 2020-08-18 12:52:42 -0700
  • fc8b8afb6f Fix a silly bug in Sample version of EvoNorm missing x* part of swish, update EvoNormBatch to accumulated unbiased variance. Ross Wightman 2020-08-13 18:23:50 -0700
  • fa26f6c1cf
         Update README.md v0.2.1 Ross Wightman 2020-08-12 23:26:59 -0700
  • f614df7921 Bump version to 0.2.1 and update README Ross Wightman 2020-08-12 18:04:08 -0700
  • b423bc8362
         Merge pull request #218 from rwightman/cutmix Ross Wightman 2020-08-12 17:17:41 -0700
  • 8c9814e3f5 Final cleanup of mixup/cutmix. Element/batch modes working with both collate (prefetcher active) and without prefetcher. #218 Ross Wightman 2020-08-12 17:01:32 -0700
  • 0f5d9d8166 Add CSPResNet50 weights, 79.6 top-1 at 256x256 Ross Wightman 2020-08-12 11:20:11 -0700
  • 0734c0d187 Update test workflow Ross Wightman 2020-08-11 19:04:37 -0700
  • b1b6e7c361 Fix a few more issues related to #216 w/ TResNet (space2depth) and FP16 weights in wide resnets. Also don't completely dump pretrained weights in in_chans != 1 or 3 cases. Ross Wightman 2020-08-11 18:57:47 -0700
  • 512b2dd645 Add new EfficientNet-B3 and RegNetY-3.2GF weights, both just over 82 top-1 Ross Wightman 2020-08-11 14:18:51 -0700
  • 6890300877 Add DropPath (stochastic depth) to RegNet Ross Wightman 2020-08-11 14:08:53 -0700
  • 47794d2c59
         Merge pull request #216 from yu4u/fix_default_cfgs Ross Wightman 2020-08-11 12:34:19 -0700
  • cd23f55397 Fix mixed prec issues with new mixup code Ross Wightman 2020-08-11 12:17:43 -0700
  • e717250ed4 remove comments #216 Yusuke Uchida 2020-08-11 23:33:22 +0900
  • f6b56602f9 fix test_model_default_cfgs Yusuke Uchida 2020-08-11 23:23:57 +0900
  • f471c17c9d More cutmix/mixup overhaul, ready to kick-off some trials. Ross Wightman 2020-08-11 00:10:33 -0700
  • 078a51dbac
         Merge pull request #214 from MohamedAliRashad/patch-1 Ross Wightman 2020-08-08 16:15:22 -0700
  • d5145fa4d5 Change default_cfg names for senet to include the legacy and match model names Ross Wightman 2020-08-08 09:01:46 -0700
  • 44d8ecc6e1
         mobilenetv3_100 doesn't exist #214 Mohamed Rashad 2020-08-08 14:10:40 +0200
  • 92f2d0d65d Merge branch 'master' into cutmix. Fixup a few issues. Ross Wightman 2020-08-07 15:59:52 -0700
  • 417b8a7bb7
         Add a warning #213 Santiago Castro 2020-08-07 17:08:50 -0400
  • 2d0ff2f444
         Fix MultiEpochsDataLoader when not all is consumed Santiago Castro 2020-08-07 16:28:48 -0400
  • 0f75d48a66
         Fix class name #212 Santiago Castro 2020-08-07 16:26:00 -0400
  • feaa3abc51
         Fix MultiEpochsDataLoader when there's no batching Santiago Castro 2020-08-07 16:25:14 -0400
  • 6e9d6172c8 A few typos and missed updates in changelog Ross Wightman 2020-08-05 22:19:59 -0700
  • 26b3a56a0a Deployed 57510fd with MkDocs version: 1.1.2 Ross Wightman 2020-08-05 16:28:17 -0800
  • 57510fd5b2 Fix some documentation rendering issues Ross Wightman 2020-08-05 17:26:55 -0700
  • 7ec600e0f5 Deployed 80c3051 with MkDocs version: 1.1.2 Ross Wightman 2020-08-05 16:26:29 -0800
  • 4f6e4af795 Deployed 80c3051 with MkDocs version: 1.1.2 Ross Wightman 2020-08-05 16:09:09 -0800
  • 80c3051f5d
         Merge pull request #175 from rwightman/features Ross Wightman 2020-08-05 17:00:43 -0700
  • 1a15502fa4 mkdocs sortable tables.js #175 Ross Wightman 2020-08-05 16:57:13 -0700
  • 1696499ce5 Bump version to 0.2.0, ready to roll (I think) Ross Wightman 2020-08-05 16:55:18 -0700
  • e62758cf4f More documentation updates, fix a typo Ross Wightman 2020-08-05 15:59:31 -0700
  • 5e333b899f README / doc tweaks Ross Wightman 2020-08-05 13:17:39 -0700
  • dfe80414a6 Add bool arg helper Ross Wightman 2020-08-05 13:17:23 -0700
  • fa28067704 Add more augmentation arguments, including a no_aug disable flag. Fix #209 Ross Wightman 2020-08-05 13:16:44 -0700
  • e3f58fc90c Continuing README and documentation updates Ross Wightman 2020-08-05 10:38:35 -0700
  • e3c11a36dc Start updating README and docs Ross Wightman 2020-08-04 16:49:03 -0700
  • b9f6d3ad18
         Update README.md #208 Corey Levinson 2020-08-04 02:04:48 -0400
  • b1f1a54de9 More uniform treatment of classifiers across all models, reduce code duplication. Ross Wightman 2020-08-03 17:21:26 -0700
  • 9806f3e1ff autosquash github workflow didn't work out, removing Ross Wightman 2020-08-02 22:22:28 -0700
  • b496b7bde9 Re-ran batch validation on all models across all sets Ross Wightman 2020-07-31 18:15:55 -0700
  • ad150e7018 Update results csv file rank/diff script and small validate script tweak for batch validation Ross Wightman 2020-07-31 17:51:22 -0700
  • d72ddafe56 Fix some checkpoint / model str regressions Ross Wightman 2020-07-29 19:43:01 -0700
  • ecdeb470f2 Cleanup real labels in validate.py, no need to track original, will do that via exteran script Ross Wightman 2020-07-29 11:15:47 -0700
  • ac18adb9c3 Remove debug print from RexNet Ross Wightman 2020-07-29 11:15:19 -0700
  • c53ec33ae0 Add synset/label indices for results generation. Add 'valid labels' to validation script to support imagenet-a/r label subsets properly. Ross Wightman 2020-07-29 00:58:57 -0700
  • ec4976fdba Add EfficientNet-Lite0 weights trained with this code by @hal-314, 75.484 top-1 Ross Wightman 2020-07-29 00:32:08 -0700
  • 9ecd16bd7b Add new seresnet50 (non-legacy) model weights, 80.274 top-1 Ross Wightman 2020-07-29 00:17:42 -0700
  • 4805dd1e79 Fix incorrect dataset name, add new datasets to results delta script Ross Wightman 2020-07-27 19:06:02 -0700
  • ea58e0bc16 Disable big models for MacOS test since they are starting to fail frequently. Ross Wightman 2020-07-27 18:01:43 -0700
  • 7995295968 Merge branch 'logger' into features. Change 'logger' to '_logger'. Ross Wightman 2020-07-27 18:00:46 -0700
  • 1998bd3180 Merge branch 'feature/AB/logger' of https://github.com/antoinebrl/pytorch-image-models into logger Ross Wightman 2020-07-27 16:06:01 -0700
  • 6be878c22a Update results README to include robustness and real labels tests. Include labels files used in results gen. Ross Wightman 2020-07-27 14:08:44 -0700
  • 6c17d57a2c Fix some attributions, add copyrights to some file docstrings Ross Wightman 2020-07-27 13:44:56 -0700
  • a69c0e04f0 Fix pool size in cspnet Ross Wightman 2020-07-27 13:44:02 -0700
  • 14ef7a0dd6 Rename csp.py -> cspnet.py Ross Wightman 2020-07-27 11:15:07 -0700
  • ec37008432 Add pretrained weight links to CSPNet for cspdarknet53, cspresnext50 Ross Wightman 2020-07-27 11:13:21 -0700
  • 78e1b56f16 Added train run for efficient net B2 #199 Jacqueline Yau 2020-07-25 21:18:42 -0700
  • e93e571f7a Add `adamp` and 'sgdp' optimizers. Sangdoo Yun 2020-07-22 11:02:05 +0900
  • 0915beddad Add autosquash workflow Ross Wightman 2020-07-25 14:26:30 -0700
  • ec9e7533a1
         Update adamp.py #195 Sangdoo Yun 2020-07-25 12:46:19 +0900
  • 0e4bbb3bba
         Update sgdp.py Sangdoo Yun 2020-07-25 12:45:25 +0900
  • ec8836166e
         Update sgdp.py Sangdoo Yun 2020-07-25 12:43:51 +0900
  • fceb09a44c
         Update adamp.py Sangdoo Yun 2020-07-25 12:43:29 +0900