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

Commit Graph

  • 6d90fcf282 Fix distribute_bn and model_ema Ross Wightman 2021-05-18 11:34:31 -0700
  • 00548b8427 Add Twins 李鑫杰 2021-05-18 19:21:53 +0800
  • 74d2829341 Merge branch 'master' into bits_and_tpu Ross Wightman 2021-05-17 16:23:14 -0700
  • aa92d7b1c5 Major timm.bits update. Updater and DeviceEnv now dataclasses, after_step closure used, metrics base impl w/ distributed reduce, many tweaks/fixes. Ross Wightman 2021-05-17 15:56:24 -0700
  • e7f0db8664 Fix drop/drop_path arg on MLP-Mixer model. Fix #641 v0.4.9 Ross Wightman 2021-05-16 08:31:52 -0700
  • 9a3ae97311 Another set of byoanet models w/ ECA channel + SA + groups Ross Wightman 2021-05-15 22:56:51 -0700
  • d53e91218e Fix tf.data options setting for newer TF versions Ross Wightman 2021-05-15 22:56:12 -0700
  • 7077f16c6a Change 21k model naming from _21k to _in21k for consistency with existing 21k models. Ross Wightman 2021-05-15 12:42:26 -0700
  • 94d4b53352 Add temporary default_cfgs to visformer models so they pass tests Ross Wightman 2021-05-15 08:41:31 -0700
  • 2a72d38ba2
         Update README.md Ross Wightman 2021-05-14 23:06:49 -0700
  • 3bffc701f1 Merge branch 'master' into levit_visformer_rednet Ross Wightman 2021-05-14 23:02:12 -0700
  • 758802e1f6
         Merge pull request #636 from rwightman/effnetv2_official Ross Wightman 2021-05-14 17:24:11 -0700
  • ecc7552c5c Add levit, levit_c, and visformer model defs. Largely untested and not finished cleanup. Ross Wightman 2021-05-14 17:10:43 -0700
  • 165fb354b2 Add initial RedNet model / Involution layer impl for testing Ross Wightman 2021-05-14 16:48:58 -0700
  • 328249f11a Update README, tweak fine-tune effv2 model names. #636 Ross Wightman 2021-05-14 16:37:43 -0700
  • c4f482a08b EfficientNetV2 official impl w/ weights ported from TF. Cleanup/refactor of related EfficientNet classes and models. Ross Wightman 2021-05-14 15:50:00 -0700
  • 49956c7674
         Merge f8ce54893c into c16d65a8a7 #603 Rahul Somani 2021-05-11 23:43:32 +0530
  • f8ce54893c add aa stem logic, minimal docstrings Rahul Somani 2021-05-11 23:43:23 +0530
  • c16d65a8a7 Update results.csv files with cait, mlp-mixer, resnetrs, miil imagenet21k pretrained, and my latest efficientnet v2s/b4 weights. v0.1-effv2-weights Ross Wightman 2021-05-11 10:50:46 -0700
  • 4fbc32d3d0 Fix crop_pct for cait models. Ross Wightman 2021-05-11 10:50:15 -0700
  • c4bb8c7d4a add model url Rahul Somani 2021-05-10 13:41:04 +0530
  • d7e1e7144a fix: the exeption when topk is 1 #626 #625 comar 2021-05-10 16:52:40 +0900
  • ffa90e04d3 fix: the exception not using default topk argument comar 2021-05-10 16:20:19 +0900
  • 715519a5ef Rethink name of patch embed grid info Ross Wightman 2021-05-06 14:08:20 -0700
  • b2c305c2aa Move Mlp and PatchEmbed modules into layers. Being used in lots of models now... Ross Wightman 2021-05-06 14:03:23 -0700
  • 3ba6b55cb2 More adjustments to ByoaNet models for further experiments. Ross Wightman 2021-05-06 13:59:44 -0700
  • f480734ad5 Update README.md Ross Wightman 2021-05-05 23:17:35 -0700
  • 504fef4bc1
         Merge pull request #609 from rwightman/cait Ross Wightman 2021-05-05 18:00:22 -0700
  • d45e50b9db Update test for cait 448x448 model #609 Ross Wightman 2021-05-05 17:51:23 -0700
  • 5fcddb96a8 Merge branch 'master' into cait Ross Wightman 2021-05-05 17:29:38 -0700
  • 3db12b4b6a Finish CaiT cleanup Ross Wightman 2021-05-05 17:28:19 -0700
  • 936e9b331d
         Merge pull request #613 from rwightman/mlp-mixer Ross Wightman 2021-05-05 16:59:29 -0700
  • 2d8b09fe8b Add official pretrained weights to MLP-Mixer, complete model cfgs. #613 Ross Wightman 2021-05-05 15:59:40 -0700
  • 12efffa6b1 Initial MLP-Mixer attempt... Ross Wightman 2021-05-05 00:59:45 -0700
  • 071affa5d9 Fixed default values in argument help messages #612 regonn 2021-05-05 15:20:49 +0900
  • 0721559511 Improved (hopefully) init for SA/SA-like layers used in ByoaNets Ross Wightman 2021-05-04 21:40:39 -0700
  • d5473c17f7 Fix incorrect name of shortcut/identity paths in many residual nets. Inherited from naming in old old torchvision, long fixed there. Ross Wightman 2021-05-04 21:27:15 -0700
  • 0d87650fea Remove filter hack from BlurPool w/ non-persistent buffer. Use BlurPool2d instead of AntiAliasing.. for TResNet. Breaks PyTorch < 1.6. Ross Wightman 2021-05-04 16:56:28 -0700
  • ddc743fdf8 Update ResNet-RS models to EMA weights Ross Wightman 2021-05-04 16:16:55 -0700
  • 08d60f4a9a resnetrs50 pool sizing wrong Ross Wightman 2021-05-04 12:41:09 -0700
  • 137a374930
         Merge pull request #555 from MichaelMonashev/patch-1 Ross Wightman 2021-05-04 11:44:01 -0700
  • 1daa15ecc3 Initial Cait commit. Still some cleanup to do. Ross Wightman 2021-05-04 11:19:15 -0700
  • 67d0665b46 Post ResNet-RS merge cleanup. Add weight urls, adjust train/test/crop pct. Ross Wightman 2021-05-04 11:04:23 -0700
  • 560eae38f5
         [WIP] Add ResNet-RS models (#554) Aman Arora 2021-05-05 03:59:44 +1000
  • 1f45c6e926 There is no bias in replaced maxpool so remove it #554 Aman Arora 2021-05-03 07:42:14 +0000
  • c7b40414bb Make sure reduction ratio in resnetrs is 0.25 Aman Arora 2021-05-03 01:49:02 +0000
  • 69f8c7123f Remove hardcoded values Aman Arora 2021-05-02 01:40:06 +0000
  • e7df94a760
         Merge branch 'master' into resnet-rs Aman Arora 2021-05-02 11:27:04 +1000
  • d9551a2019 Back out temporary code #580 Sander van Leeuwen 2021-05-01 15:34:29 +0200
  • acc0d37d96 Test script for parsing image files Sander van Leeuwen 2021-05-01 15:33:52 +0200
  • e060223cee add default configs for aa models Rahul Somani 2021-05-01 12:10:49 +0530
  • c2abb2c03d add anti-aliasing for `ir` block, mobnet-v3 Rahul Somani 2021-05-01 11:50:43 +0530
  • 9cc7dda6e5 Fixup byoanet configs to pass unit tests. Add swin_attn and swinnet26t model for testing. v0.1-rs-weights Ross Wightman 2021-04-29 21:08:37 -0700
  • e15c3886ba Defaul lambda r=7. Define '26t' stage 4/5 256x256 variants for all of bot/halo/lambda nets for experiment. Add resnet50t for exp. Fix a few comments. Ross Wightman 2021-04-29 10:58:49 -0700
  • d15ad3e919 Merge branch 'morizin-master' Ross Wightman 2021-04-28 18:10:12 -0700
  • e5e15754c9 Fix coat first conv ident Ross Wightman 2021-04-28 18:09:23 -0700
  • 76739a7589 CoaT merge. Bit of formatting, fix torchscript (for non features), remove einops/einsum dep, add pretrained weight hub (url) support. Ross Wightman 2021-04-28 16:31:35 -0700
  • 026430c083 Merge branch 'master' of https://github.com/morizin/pytorch-image-models-1 into morizin-master Ross Wightman 2021-04-28 11:24:27 -0700
  • c03e45cbf3 Merge branch 'mrT23-master' v0.1-coat-weights Ross Wightman 2021-04-28 10:33:48 -0700
  • a0492e3b48 A few miil weights naming tweaks to improve compat with model registry and filtering wildcards. Ross Wightman 2021-04-28 10:33:02 -0700
  • 8c1f03e56c comment #595 talrid 2021-04-28 09:04:10 +0300
  • 19e1b67a84 old spaces talrid 2021-04-28 09:01:29 +0300
  • a443865876 update naming and scores talrid 2021-04-28 08:57:06 +0300
  • cf0e371594 84_0 talrid 2021-04-27 22:33:55 +0300
  • 0968bdeca3 vit, tresnet and mobilenetV3 ImageNet-21K-P weights talrid 2021-04-27 20:54:06 +0300
  • b81cd75eea
         Merge pull request #3 from rwightman/master mrT23 2021-04-27 18:49:25 +0300
  • 1e3b6d4dfc
         Update __init__.py #581 morizin 2021-04-24 18:22:59 +0530
  • fd022fd6a2
         Update __init__.py morizin 2021-04-24 18:22:36 +0530
  • 06841427cd
         Add files via upload morizin 2021-04-24 18:11:28 +0530
  • c2d5087eae
         Add files via upload morizin 2021-04-24 17:47:57 +0530
  • 7afadae610 Adjust rate calculation for different batch sizes Sander van Leeuwen 2021-04-23 13:18:17 +0200
  • a9699d9e45 Parameter name change Sander van Leeuwen 2021-04-23 11:36:19 +0200
  • 7de69cd325 Added support for distributing workload among GPUs of diffrent generations (speed/memory) Sander van Leeuwen 2021-04-23 11:34:33 +0200
  • 938716c753 Fix import issue, use devenv for dist info in parser_tfds Ross Wightman 2021-04-21 13:16:11 -0700
  • 76de984a5f Fix some bugs with XLA support, logger, add hacky xla dist launch script since torch.dist.launch doesn't work Ross Wightman 2021-04-21 13:02:53 -0700
  • 12d9a6d4d2 First timm.bits commit, add initial abstractions, WIP updates to train, val... some of it working Ross Wightman 2021-04-20 17:12:02 -0700
  • 9a1bd358c7
         Merge pull request #571 from normster/augmix-fix Ross Wightman 2021-04-19 16:18:36 -0700
  • 79640fcc1f Enable uniform augmentation magnitude sampling and set AugMix default #571 Norman Mu 2021-04-19 14:19:31 -0700
  • c1cf9712fc Add updated EfficientNet-V2S weights, 83.8 @ 384x384 test. Add PyTorch trained EfficientNet-B4 weights, 83.4 @ 384x384 test. Tweak non TF EfficientNet B1-B4 train/test res scaling. Ross Wightman 2021-04-19 10:42:56 -0700
  • e8a64fb881 Test input size for efficientnet_v2s was wrong in last results run Ross Wightman 2021-04-17 16:17:41 -0700
  • a04427d8ce Add _in22k to bulk validate filter Ross Wightman 2021-04-17 12:17:45 -0700
  • af647c10b3 Update results csv, includes latest transformer models, swin, pit, tnt... run on pytorch 1.8.1 cuda 10.2 (11.x rel crashes). Ross Wightman 2021-04-17 12:16:05 -0700
  • e15e68d881 Fix #566, summary.csv writing to pwd on local_rank != 0. Tweak benchmark mem handling to see if it reduces likelihood of 'bad' exceptions on OOM. Ross Wightman 2021-04-15 23:03:56 -0700
  • 1b0c8e7b01 Merge branch 'iamhankai-master' Ross Wightman 2021-04-15 10:22:42 -0700
  • 2df77ee5cb Fix torchscript compat and features_only behaviour in GhostNet PR. A few minor formatting changes. Reuse existing layers. Ross Wightman 2021-04-15 10:20:26 -0700
  • d793deb51a Merge branch 'master' of https://github.com/iamhankai/pytorch-image-models into iamhankai-master Ross Wightman 2021-04-15 09:30:25 -0700
  • e685618f45
         Merge pull request #550 from amaarora/wandb Ross Wightman 2021-04-15 09:26:35 -0700
  • 277a9a78f9 Fix unit test filter update. Ross Wightman 2021-04-13 22:15:20 -0700
  • 858728799c Update README again. Add 101x3 BiT-M model to CI ignore since it's starting to fail in GitHub runners. Ross Wightman 2021-04-13 13:36:24 -0700
  • f606c45c38 Add Swin Transformer models from https://github.com/microsoft/Swin-Transformer Ross Wightman 2021-04-13 12:17:21 -0700
  • de445e7827 Add GhostNet #561 iamhankai 2021-04-07 22:51:15 +0800
  • 5a196dddf6 Update README.md with latest, bump version to 0.4.8 Ross Wightman 2021-04-12 13:15:00 -0700
  • ce6585f533
         Merge pull request #556 from rwightman/byoanet-self_attn Ross Wightman 2021-04-12 13:04:24 -0700
  • b3d7580df1 Update ByoaNet comments. Fix first Steam feat chs for ByobNet. #556 Ross Wightman 2021-04-12 12:11:35 -0700
  • 16f7aa9f54 Add default_cfg options for min_input_size / fixed_input_size, queries in model registry, and use for testing self-attn models Ross Wightman 2021-04-12 11:54:22 -0700
  • ea9f935e9d Remove resnetv2_101x1_bitm from exclude feat features Aman Arora 2021-04-12 14:18:43 -0400
  • 3a73401b5d Add bigger resnetrs to model filters to fix Github CLI Aman Arora 2021-04-12 14:17:14 -0400
  • 4e4b863b15 Missed norm.py Ross Wightman 2021-04-12 09:57:56 -0700
  • 7c97e66f7c Remove commented code, add more consistent seed fn Ross Wightman 2021-04-12 09:51:36 -0700
  • 364dd6a58e Merge branch 'master' into byoanet-self_attn Ross Wightman 2021-04-12 09:38:59 -0700