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

Commit Graph

  • 7d657d2ef4 Improve resolve_pretrained_cfg behaviour when no cfg exists, warn instead of crash. Improve usability ex #1311 Ross Wightman 2022-06-24 14:55:25 -0700
  • 879df47c0a Support BatchNormAct2d for sync-bn use. Fix #1254 Ross Wightman 2022-06-24 14:51:26 -0700
  • 7cedc8d474 Follow up to #1256, fix interpolation warning in auto_autoaugment as well Ross Wightman 2022-06-21 14:56:53 -0700
  • 037e5e6c09 Fix #1309, move wandb init after distributed init, only init on rank == 0 process Ross Wightman 2022-06-21 12:31:48 -0700
  • 9e12530433 use utils namespace instead of function/classnames Jakub Kaczmarzyk 2022-05-26 08:57:47 -0400
  • db64393c0d
         use `Image.Resampling` namespace for PIL mapping (#1256) Jakub Kaczmarzyk 2022-06-13 01:30:57 -0400
  • cf57f95b0e use new pillow resampling enum only if it exists #1256 Jakub Kaczmarzyk 2022-06-12 15:09:34 -0400
  • cf57377c82
         Merge 1af13ab740 into db8e33c69f #1213 Shuangchi He 2022-06-11 17:46:40 +0800
  • db8e33c69f
         Merge pull request #1294 from xwang233/add-aot-autograd Ross Wightman 2022-06-09 14:32:11 -0700
  • 2d7ab06503 Move aot-autograd opt after model metadata used to setup data config in benchmark.py #1294 Ross Wightman 2022-06-09 14:30:21 -0700
  • ca991c1fa5 add --aot-autograd Xiao Wang 2022-06-07 18:01:52 -0700
  • 1af13ab740
         Update changes.md #1213 Shuangchi He 2022-05-28 18:46:38 +0800
  • 2b2196028f
         Add missing output #1277 lukasugar 2022-05-26 19:23:23 +0200
  • 0f25d1838c use utils namespace instead of function/classnames #1276 Jakub Kaczmarzyk 2022-05-26 08:57:47 -0400
  • e7ae390a77 Add tif support #1274 Kyriaki Dionysopoulou 2022-05-26 07:40:02 +0000
  • 99205d7be5
         Revert "ENH: replace star imports with imported names in train.py" #1273 Jakub Kaczmarzyk 2022-05-25 19:10:34 -0400
  • dff33730b3 Merge remote-tracking branch 'origin/master' into bits_and_tpu Ross Wightman 2022-05-25 11:27:07 -0700
  • 089d9046bf
         Merge 01010b5aa7 into e4360e6125 #1271 X Wang 2022-05-24 18:53:35 +0000
  • e4360e6125
         Merge pull request #1270 from developer0hye/patch-1 Ross Wightman 2022-05-24 10:40:23 -0700
  • 01010b5aa7 add some comments #1271 Xiao Wang 2022-05-24 10:36:39 -0700
  • be54f860d0 update cuda graph mode for AMP and DDP Xiao Wang 2022-05-24 10:26:26 -0700
  • 9b299cbcf1 get static data on a side stream Xiao Wang 2022-05-24 10:01:26 -0700
  • cd926775fb added aot_autograd; cuda graph still wip Xiao Wang 2022-05-24 09:40:58 -0700
  • 1726f27f8d add cuda graph and aot_autograd to benchmark.py Xiao Wang 2022-05-24 09:40:46 -0700
  • 57f8361a01
         fix a function parameter typo(cropt_pct -> crop_pct) #1270 Yonghye Kwon 2022-05-25 00:36:28 +0900
  • fd360ac951
         Merge pull request #1266 from kaczmarj/enh/no-star-imports Ross Wightman 2022-05-20 08:55:07 -0700
  • 004fb5f8c4
         Merge pull request #1265 from kaczmarj/enh/argparse-groups Ross Wightman 2022-05-20 08:54:36 -0700
  • ce5578bc3a replace star imports with imported names #1266 Jakub Kaczmarzyk 2022-05-18 11:04:10 -0400
  • dcad288fd6 use argparse groups to group arguments #1265 Jakub Kaczmarzyk 2022-05-18 10:27:33 -0400
  • e1e4c9bbae rm whitespace Jakub Kaczmarzyk 2022-05-18 10:17:02 -0400
  • 20a1fa63f8 Make dev version 0.6.2.dev0 for pypi pre Ross Wightman 2022-05-15 14:29:57 -0700
  • 3409bb9156
         Update README.md Ross Wightman 2022-05-13 15:28:46 -0700
  • 137292644b
         Update README.md Ross Wightman 2022-05-13 15:28:04 -0700
  • 14f370ca1d
         Merge pull request #1259 from rwightman/swin_v2 Ross Wightman 2022-05-13 15:10:15 -0700
  • 347308faad Update README.md, version to 0.6.2 #1259 swin_v2 Ross Wightman 2022-05-13 13:54:41 -0700
  • 4b30bae67b Add updated vit_relpos weights, and impl w/ support for official swin-v2 differences for relpos. Add bias control support for MLP layers Ross Wightman 2022-05-13 13:53:57 -0700
  • d4c0588012 Remove persistent buffers from Swin-V2. Change SwinV2Cr cos attn + tau/logit_scale to match official, add ckpt convert, init_value zeros resid LN weight by default Ross Wightman 2022-05-13 10:49:11 -0700
  • 27c42f0830 Fix torchscript use for offician Swin-V2, add support for non-square window/shift to WindowAttn/Block Ross Wightman 2022-05-13 09:29:33 -0700
  • 2f2b22d8c7 Disable nvfuser fma / opt level overrides per #1244 Ross Wightman 2022-05-13 09:26:16 -0700
  • c0211b0bf7 Swin-V2 test fixes, typo Ross Wightman 2022-05-12 22:31:55 -0700
  • 9a86b900fa Official SwinV2 models Ross Wightman 2022-05-12 15:01:23 -0700
  • 19b65a5c27
         use `Image.Resampling` namespace for PIL mapping Jakub Kaczmarzyk 2022-05-11 15:06:42 -0400
  • d07d015173
         Merge pull request #1249 from okojoalg/sequencer Ross Wightman 2022-05-09 20:42:43 -0700
  • d30685c283
         Merge pull request #1251 from hankyul2/fix-multistep-scheduler Ross Wightman 2022-05-09 16:07:46 -0700
  • a16171335b fix: change milestones to decay-milestones #1251 han 2022-05-10 07:57:19 +0900
  • 39b725e1c9 Fix tests for rank-4 output where feature channels dim is -1 (3) and not 1 #1249 Ross Wightman 2022-05-09 15:20:24 -0700
  • 78a32655fa Fix poolformer group_matcher to merge proj downsample with previous block, support coarse Ross Wightman 2022-05-09 12:20:04 -0700
  • d79f3d9d1e Fix torchscript use for sequencer, add group_matcher, forward_head support, minor formatting Ross Wightman 2022-05-09 12:09:39 -0700
  • 37b6920df3 Fix group_matcher regex for regnet.py Ross Wightman 2022-05-09 10:40:40 -0700
  • 4bbd9a0977
         Add pandas as a required package #1253 Mohamed Rashad 2022-05-09 07:34:44 +0200
  • cca4e94a82
         Forgot to import pandas ^^ Mohamed Rashad 2022-05-09 04:35:01 +0200
  • 23aeb88b57
         Add `list_benchmarks` to compare models easily Mohamed Rashad 2022-05-06 23:03:07 +0200
  • 93a79a3dd9 Fix num_features in Sequencer okojoalg 2022-05-06 23:16:32 +0900
  • 2fec08e923 Add Sequencer to non std filters okojoalg 2022-05-06 23:08:10 +0900
  • 57a988df30 fix: multistep lr decay epoch bugs han 2022-05-06 13:14:43 +0900
  • 578d52e752 Add Sequencer okojoalg 2022-05-05 23:22:40 +0900
  • 6d4665bb52
         Merge pull request #1245 from rwightman/vit_relpos_refactor Ross Wightman 2022-05-03 09:27:02 -0700
  • f5ca4141f7 Adjust arg order for recent vit model args, add a few comments #1245 vit_relpos_refactor Ross Wightman 2022-05-02 22:41:38 -0700
  • 41dc49a337 Vision Transformer refactoring and Rel Pos impl Ross Wightman 2022-05-02 15:37:39 -0700
  • 9c321be330
         Merge pull request #1239 from dedeswim/parser-fix Ross Wightman 2022-04-28 07:43:15 -0700
  • c76d772670 Add support for different TFDS `BuilderConfig`s #1239 Edoardo Debenedetti 2022-04-28 10:04:03 +0000
  • b7cb8d0337 Add Swin-V2 Small-NS weights (83.5 @ 224). Add layer scale like 'init_values' via post-norm LN weight scaling Ross Wightman 2022-04-26 17:32:49 -0700
  • 001688dabf
         Merge pull request #1233 from jjsjann123/nhwc_cond_conv2d Ross Wightman 2022-04-25 20:42:03 -0700
  • f88c606fcf fixing channels_last on cond_conv2d; update nvfuser debug env variable #1233 jjsjann123 2022-04-25 12:41:46 -0700
  • 7d235c5a5f
         Merge pull request #1230 from donglixp/patch-1 Ross Wightman 2022-04-24 14:06:26 -0700
  • 09e9f3defb
         migrate azure blob for beit checkpoints #1230 Li Dong 2022-04-23 13:02:29 +0800
  • 52ac881402 Missed first_conv in latest seresnext 'D' default_cfgs Ross Wightman 2022-04-22 20:55:52 -0700
  • a4adf73213 Deployed 7629d82 with MkDocs version: 1.3.0 Ross Wightman 2022-04-22 17:00:59 -0700
  • 7629d8264d Add two new SE-ResNeXt101-D 32x8d weights, one anti-aliased and one not. Reshuffle default_cfgs vs model entrypoints for resnet.py so they are better aligned. Ross Wightman 2022-04-22 16:54:53 -0700
  • fbf597049c Update README and change timmdocs link in documentation Ross Wightman 2022-04-22 16:52:05 -0700
  • e940755778 added new wt initializtion- xavier_uniform #1229 naman jain 2022-04-22 03:16:13 -0500
  • 450b0d57f0 model_freeze_and_pretraining kira7005 2022-04-21 17:06:32 +0000
  • 50fd5be983 add_softmax_layer kira7005 2022-04-20 06:06:21 +0000
  • cbc75c808e model changes to support pretrained weights loading kira7005 2022-04-20 01:40:45 +0000
  • 569d114ef7
         Fix device problem #1222 Junming Chen 2022-04-19 11:53:18 +0800
  • 592e2984af classes_number kira7005 2022-04-14 18:19:19 +0000
  • af21ef0847 bag_sampler kira7005 2022-04-12 22:52:22 +0000
  • 029a2cb449 mil_ranking loss kira7005 2022-04-12 06:34:21 +0000
  • 2f7859f887 Fix typos. Yulv-git 2022-04-09 16:30:14 +0800
  • bab39cf7b7 fix(convnext): stride warning #1212 xiejisong 2022-04-07 15:44:27 +0800
  • 2be436cb2c
         Remove dead code return #1210 Daniel Nouri 2022-04-07 08:52:17 +0200
  • c16fec9252 In load_pretrained helper: Fail if cannot find pretrained model Daniel Nouri 2022-04-06 10:24:12 +0200
  • 01a0e25a67
         Merge pull request #1208 from seefun/master Ross Wightman 2022-04-05 14:12:31 -0700
  • 8f0bc0591e fix convnext args #1208 SeeFun 2022-04-05 20:00:57 +0800
  • b0d2fcf647
         Merge branch 'rwightman:master' into master SeeFun 2022-04-05 19:23:18 +0800
  • 754e11402a Merge remote-tracking branch 'origin/master' into bits_and_tpu Ross Wightman 2022-04-04 16:23:24 -0700
  • eac2df3d2c Update PyTorch 1.10 benchmark numbers for latest code Ross Wightman 2022-04-04 09:02:54 -0700
  • c5a8e929fb Add initial swinv2 tiny / small weights Ross Wightman 2022-04-03 15:22:55 -0700
  • 83d7a11eec Update .gitignore, remove out of date notebooks v0.1-weights-swinv2 Ross Wightman 2022-04-03 14:31:26 -0700
  • 02b806e00a Add PyTorch 1.11 train benchmark numbers Ross Wightman 2022-04-01 16:34:23 -0700
  • c9f208f7f8 Update eval results and add latest PyTorch 1.11 inference benchmarks Ross Wightman 2022-03-29 16:36:31 -0700
  • 46e25ba673 Optimize the performance of ConvNeXt models #1194 Ceshine Lee 2022-03-28 17:21:31 +0800
  • 6c376d8139 Supports Read from c3d_embeddings kira7005 2022-03-27 23:31:51 +0000
  • 1ba0ec4c18 Merge remote-tracking branch 'origin/master' into bits_and_tpu Ross Wightman 2022-03-25 17:45:31 -0700
  • f670d98cb8 Make a few more layers symbolically traceable (remove from FX leaf modules) * remove dtype kwarg from .to() calls in EvoNorm as it messed up script + trace combo * BatchNormAct2d always uses custom forward (cut & paste from original) instead of super().forward. Fixes #1176 * BlurPool groups==channels, no need to use input.dim[1] Ross Wightman 2022-03-24 21:40:34 -0700
  • a9ecb880e5
         Merge pull request #1190 from seefun/ConvNeXt-pretrain Ross Wightman 2022-03-24 13:49:16 -0700
  • 5f4de2334b
         Merge pull request #1 from seefun/ConvNeXt-pretrain SeeFun 2022-03-24 15:19:49 +0800
  • ec4e9aa5a0
         Add ConvNeXt tiny and small pretrain in22k #1190 SeeFun 2022-03-24 15:18:08 +0800
  • 73ffade1f8
         Update README.md Ross Wightman 2022-03-23 22:00:31 -0700
  • 575924ed60 Update test crop for new RegNet-V weights to match Y Ross Wightman 2022-03-23 21:40:53 -0700