Commit Graph

810 Commits (f606c45c3860bbff297b3a2a30e5d574674f3179)
 

Author SHA1 Message Date
Ross Wightman f606c45c38 Add Swin Transformer models from https://github.com/microsoft/Swin-Transformer 2021-04-13 12:17:21 -07:00
Ross Wightman 5a196dddf6 Update README.md with latest, bump version to 0.4.8 2021-04-12 13:15:00 -07:00
Ross Wightman ce6585f533
Merge pull request #556 from rwightman/byoanet-self_attn
ByoaNet - Self Attn Networks - Bottleneck Transformers, Lambda ResNet, HaloNet
2021-04-12 13:04:24 -07:00
Ross Wightman b3d7580df1 Update ByoaNet comments. Fix first Steam feat chs for ByobNet. 2021-04-12 12:11:35 -07:00
Ross Wightman 16f7aa9f54 Add default_cfg options for min_input_size / fixed_input_size, queries in model registry, and use for testing self-attn models 2021-04-12 11:54:22 -07:00
Ross Wightman 4e4b863b15 Missed norm.py 2021-04-12 09:57:56 -07:00
Ross Wightman 7c97e66f7c Remove commented code, add more consistent seed fn 2021-04-12 09:51:36 -07:00
Ross Wightman 364dd6a58e Merge branch 'master' into byoanet-self_attn 2021-04-12 09:38:59 -07:00
Ross Wightman ce62f96d4d ByoaNet with bottleneck transformer, lambda resnet, and halo net experiments 2021-04-12 09:38:02 -07:00
Ross Wightman cd3dc4979f Fix adabelief imports, remove prints, preserve memory format is the default arg for zeros_like 2021-04-12 08:25:31 -07:00
Ross Wightman 21812d33aa Add prelim efficientnet_v2s weights from 224x224 train, eval 83.3 @ 288. Add eca_nfnet_l1 weights, train at 256, eval 84 @ 320. 2021-04-11 23:26:13 -07:00
Ross Wightman 779107b693
Merge pull request #542 from juntang-zhuang/adabelief
Add Adabelief Optimizer
2021-04-05 16:34:54 -07:00
Juntang Zhuang 74366f733c
Delete distributed_train_adabelief.sh 2021-04-05 15:11:45 -04:00
Juntang Zhuang 1d848f409a
Delete args.yaml 2021-04-05 15:11:35 -04:00
juntang addfc7c1ac adabelief 2021-04-04 23:48:15 -04:00
Ross Wightman fb896c0b26 Update some comments re preliminary EfficientNet-V2 assumptions 2021-04-03 12:00:25 -07:00
Ross Wightman 2b49ab7a36 Fix ResNetV2 pretrained classifier issue. Fixes #540 2021-04-03 11:18:12 -07:00
Ross Wightman de9dff933a EfficientNet-V2S preliminary model def (for experimentation) 2021-04-02 09:36:51 -07:00
Ross Wightman d5ed58d623
Merge pull request #533 from rwightman/pit_and_vit_update
Addition of PiT models and update/cleanup of ViT, new NFNet weight, TFDS wrapper fix, few misc fixes/updates
2021-04-01 22:56:59 -07:00
Ross Wightman 37c71a5609 Some further create_optimizer_v2 tweaks, remove some redudnant code, add back safe model str. Benchmark step times per batch. 2021-04-01 22:34:55 -07:00
Ross Wightman 2bb65bd875 Wrong default_cfg pool_size for L1 2021-04-01 20:00:41 -07:00
Ross Wightman bf2ca6bdf4 Merge jax and original weight init 2021-04-01 18:11:51 -07:00
Ross Wightman acbd698c83 Update README.md with updates. Small tweak to head_dist handling. 2021-04-01 17:49:05 -07:00
Ross Wightman 9071568f0e Add weights for SE NFNet-L0 model, rename nfnet_l0b -> nfnet_l0. 82.75 top-1 @ 288. Add nfnet_l1 model def for training. 2021-04-01 17:22:27 -07:00
Ross Wightman c468c47a9c Add regnety_160 weights from DeiT teacher model, update that and my regnety_032 weights to use higher test size. 2021-04-01 16:41:04 -07:00
Ross Wightman 288682796f Update benchmark script to add precision arg. Fix some downstream (DeiT) compat issues with latest changes. Bump version to 0.4.7 2021-04-01 16:40:12 -07:00
Ross Wightman ea9c9550b2 Fully move ViT hybrids to their own file, including embedding module. Remove some extra DeiT models that were for benchmarking only. 2021-04-01 14:17:38 -07:00
Ross Wightman a5310a3451 Merge remote-tracking branch 'origin/benchmark-fixes-vit_hybrids' into pit_and_vit_update 2021-04-01 12:15:34 -07:00
Ross Wightman 7953e5d11a Fix pos_embed scaling for ViT and num_classes != 1000 for pretrained distilled deit and pit models. Fix #426 and fix #433 2021-03-31 23:11:28 -07:00
Ross Wightman a760a4c3f4 Some ViT cleanup, merge distilled model with main, fixup torchscript support for distilled models 2021-03-31 18:21:02 -07:00
Ross Wightman 0dfc5a66bb Add PiT model from https://github.com/naver-ai/pit 2021-03-31 18:20:14 -07:00
Ross Wightman 1ad1645a50 Merge branch 'contrastive-master' 2021-03-29 11:33:32 -07:00
Ross Wightman 51febd869b Small tweak to tests for tnt model, reorder model imports. 2021-03-29 11:33:08 -07:00
Ross Wightman b27a4e0d88 Merge branch 'master' of https://github.com/contrastive/pytorch-image-models into contrastive-master 2021-03-29 10:37:05 -07:00
Ross Wightman 2319cbbff2
Merge pull request #525 from amaarora/spp
Add `ActivationStatsHook`  to allow extracting activation stats for Signal Propogation Plots
2021-03-29 08:36:04 -07:00
Aman Arora 6b18061773 Add GIST to docstring for quick access 2021-03-29 15:33:31 +11:00
contrastive 809271b0f3
Update test_models.py 2021-03-29 09:39:50 +08:00
contrastive de86314655 Update TNT 2021-03-29 08:23:34 +08:00
Aman Arora 92b1db9a79 update docstrings and add check on and 2021-03-29 10:04:51 +11:00
Aman Arora b85be24054 update to work with fnmatch 2021-03-29 09:36:31 +11:00
contrastive cfc15283a4 Update TNT url 2021-03-28 23:19:15 +08:00
contrastive 4a09bc851e Add TNT model 2021-03-28 19:53:42 +08:00
Aman Arora 20626e8387 Add to extract stats for SPP 2021-03-27 05:40:04 +11:00
Ross Wightman a2727c1bf7
Merge pull request #510 from rwightman/dependabot/pip/jinja2-2.11.3
Bump jinja2 from 2.11.2 to 2.11.3
2021-03-20 16:45:53 -07:00
Ross Wightman e2e3290fbf Add '--experiment' to train args for fixed exp name if desired, 'train' not added to output folder if specified. 2021-03-20 12:02:17 -07:00
Ross Wightman cf5fec5047 Cleanup experimental vit weight init a bit 2021-03-20 09:44:24 -07:00
dependabot[bot] b36099ace3
Bump jinja2 from 2.11.2 to 2.11.3
Bumps [jinja2](https://github.com/pallets/jinja) from 2.11.2 to 2.11.3.
- [Release notes](https://github.com/pallets/jinja/releases)
- [Changelog](https://github.com/pallets/jinja/blob/master/CHANGES.rst)
- [Commits](https://github.com/pallets/jinja/compare/2.11.2...2.11.3)

Signed-off-by: dependabot[bot] <support@github.com>
2021-03-20 05:50:25 +00:00
Ross Wightman f42f1df26c Improve evenness of per-worker split for validation set with TFDS 2021-03-18 23:16:14 -07:00
Ross Wightman cbcb76d72c Should have included Conv2d layers in original weight init. Lets see what the impact is... 2021-03-18 23:15:48 -07:00
Ross Wightman 4de57ccf01 Add weight init scheme that's closer to JAX impl 2021-03-18 15:35:22 -07:00