Commit Graph

  • 68174b9367
    Merge 3dc90ed7a7 into c8c4f256b8 Ross Wightman 2025-04-25 23:01:50 +0000
  • 3dc90ed7a7 Add naflex loader support to validate.py, fix bug in naflex pos embed add, classic vit weight loading for naflex model naflex Ross Wightman 2025-04-25 16:00:54 -0700
  • 61b7922f6b
    Merge b82f9e869c into c8c4f256b8 Bernie Huang 2025-04-25 14:48:46 -0700
  • b82f9e869c renaming models and update checkpoint to vit-only berniebear 2025-04-25 21:26:34 +0000
  • 3af564f2fb fix default config berniebear 2025-04-25 09:19:18 +0000
  • e6bbf9fd56 add default config berniebear 2025-04-25 07:55:00 +0000
  • 8eafe2c21e update pe to reuse timm layers berniebear 2025-04-25 06:50:29 +0000
  • 0f6e29019c pe integration berniebear 2025-04-24 22:39:32 +0000
  • 8e7c4a5bf1
    Merge c9c973b3b5 into c8c4f256b8 Ross Wightman 2025-04-24 21:32:51 +0200
  • f393439d2c
    Merge 82f75f4cbb into c8c4f256b8 Bill Psomas 2025-04-23 22:33:03 +0300
  • c0a54c4caa
    Merge c3445e9d75 into c8c4f256b8 Thien Tran 2025-04-23 22:31:03 +0300
  • ef03c43335
    Merge b283c24f13 into c8c4f256b8 Ryan 2025-04-23 22:42:29 +0800
  • 06309819fc
    Merge 6c00fa52b4 into c8c4f256b8 Rorry Brenner 2025-04-22 19:49:59 -0400
  • c8c4f256b8
    add-lightly-train-to-training-frameworks-readme (#2474) main Yutong Xiang 2025-04-22 17:50:11 +0200
  • c527c37969 Optimizations for pos embed resize, merge different mask helper fns Ross Wightman 2025-04-21 14:05:18 -0700
  • 88e8342297
    Merge f63d2e8071 into 0c136f7dab Fredo Guan 2025-04-21 10:47:32 +0200
  • 506f12b16e
    Merge 22b1f35dc6 into 0c136f7dab JosuaRieder 2025-04-19 16:48:06 +0000
  • 0c136f7dab fix img_size type Xihua Dong 2025-04-18 09:13:55 -0700
  • 9035c459ca fix img_size type Xihua Dong 2025-04-18 09:13:55 -0700
  • 24dbb34718
    Merge 430f7b4c5d into 3ff3899026 Fredo Guan 2025-04-18 17:13:33 +0200
  • 3ff3899026 Add local-dir: schema support for model loading (config + weights) from folder Ross Wightman 2025-04-17 10:32:48 -0700
  • fe353419af Add local-dir: schema support for model loading (config + weights) from folder local_dir Ross Wightman 2025-04-17 10:32:48 -0700
  • 7685d4f0d1
    Merge 3abb7811fe into ceca5efdec JosuaRieder 2025-04-16 06:38:35 +0800
  • ceca5efdec Remove torch_out from onnx export, no point without the export_ fn Ross Wightman 2025-04-15 12:07:19 -0700
  • 0cae8a4cd8 Fix #2472, torch.onnx.export_ (with return output) finally removed :( Ross Wightman 2025-04-15 12:03:03 -0700
  • 990f618868 Remove torch_out from onnx export, no point without the export_ fn fix_onnx Ross Wightman 2025-04-15 12:07:19 -0700
  • b1d09e2d49 Fix #2472, torch.onnx.export_ (with return output) finally removed :( Ross Wightman 2025-04-15 12:03:03 -0700
  • 71469258db
    Merge branch 'huggingface:main' into main Yutong Xiang 2025-04-15 15:24:45 +0200
  • ea728f67fa Improve several typing issues for flex vit, can (almost) work with jit if we bash h,w key into an int or str Ross Wightman 2025-04-14 11:01:56 -0700
  • 681be882e8 Fix arg merging of sknet, old seresnet. Fix #2470 Ross Wightman 2025-04-14 09:24:31 -0700
  • 382444362a Fix arg merging of sknet, old seresnet. Fix #2470 sknet_kwargs_merge Ross Wightman 2025-04-14 09:24:31 -0700
  • 97341fec51 A much faster resample_patch_embed, can be used at train/validation time Ross Wightman 2025-04-10 15:58:24 -0700
  • 62d588fcf3
    Merge pull request #1 from lightly-ai/add-lightly-train-to-training-frameworks-readme Yutong Xiang 2025-04-10 16:22:16 +0200
  • 37091f0da0 add-lightly-train-to-training-frameworks-readme yutong 2025-04-10 15:32:47 +0200
  • b4bb0f452a Exclude embeds module and mask attn functions from tracing Ross Wightman 2025-04-09 15:34:15 -0700
  • 13e0f3a4a3 Add loss scale arg, initial distributed loss scale. Maybe fix FX for the model. Ross Wightman 2025-04-08 20:47:57 -0700
  • 6675590264 Fix ParallelThingsBlock w/ attn_mask Ross Wightman 2025-04-08 09:35:34 -0700
  • 9b23d6dea2 Exclude naflex models from jit tests Ross Wightman 2025-04-08 07:59:19 -0700
  • 825edccf19 Type fixes, remove old comments Ross Wightman 2025-04-07 21:35:03 -0700
  • 0893f5d296 Initial NaFlex ViT model and training support Ross Wightman 2025-04-07 21:27:10 -0700
  • c3445e9d75 support pytorch<2 Thien Tran 2025-03-30 17:16:25 +0800
  • d70c481179 override device kwargs of base nn classes Thien Tran 2025-03-30 17:05:25 +0800
  • 6c00fa52b4 updated reader_factory to correct extra folders Rorry 2025-03-02 22:09:03 -0500
  • cdfcd3d058
    pyproject.toml: add required dep versions Adam J. Stewart 2025-02-23 09:08:06 +0100
  • e44f14d7d2 Update README v1.0.15 Ross Wightman 2025-02-22 21:04:13 -0800
  • 98e9651952
    Update version.py Ross Wightman 2025-02-22 10:50:21 -0800
  • e76ea5474d
    Update README.md Ross Wightman 2025-02-21 16:09:42 -0800
  • 92682d8d4d timm.models: explicitly export attributes Adam J. Stewart 2025-02-21 19:33:00 +0100
  • a667d3d8f0 siglip2 weights on hub, fix forward_intermediates when no prefix tokens (& return prefix selected) Ross Wightman 2025-02-21 12:46:14 -0800
  • f63a11cf81 Remove duplicate so400m/16 @ 256 model def Ross Wightman 2025-02-20 16:14:58 -0800
  • 9758e0b8b0 Prep for siglip2 release Ross Wightman 2025-02-20 12:07:49 -0800
  • 228e080e39 siglip2 weights on hub, fix forward_intermediates when no prefix tokens (& return prefix selected) siglip2 Ross Wightman 2025-02-21 12:46:14 -0800
  • c68d724e9c adapt_input_conv: add type hints Adam J. Stewart 2025-02-21 19:24:05 +0100
  • b1ee8fb96c
    timm.models: explicitly export attributes Adam J. Stewart 2025-02-21 19:33:00 +0100
  • d2cbc7fb21
    adapt_input_conv: add type hints Adam J. Stewart 2025-02-21 19:24:05 +0100
  • 25de6b4ee1 Remove duplicate so400m/16 @ 256 model def Ross Wightman 2025-02-20 16:14:58 -0800
  • dc5c7989ae Prep for siglip2 release Ross Wightman 2025-02-20 12:07:49 -0800
  • 105a667baa Dev version 1.0.15.dev0 Ross Wightman 2025-02-17 15:50:12 -0800
  • 7234f5c6c5 Add 448 so150m2 weight/model, add updated internvit 300m weight Ross Wightman 2025-02-17 12:59:10 -0800
  • 9ce824c39a Add vit so150m2 weights Ross Wightman 2025-02-14 15:02:32 -0800
  • 5942f1d492 Add vit so150m2 weights so150m2 Ross Wightman 2025-02-14 15:02:32 -0800
  • a49b020eff Merge branch 'ClashLuke-patch-1' Ross Wightman 2025-01-31 12:53:29 -0800
  • 490d222dd8 Fix issue taking device from V before V exists Ross Wightman 2025-01-31 12:52:27 -0800
  • 875c19d0c9 Merge branch 'patch-1' of github.com:ClashLuke/pytorch-image-models into ClashLuke-patch-1 Ross Wightman 2025-01-31 12:43:28 -0800
  • 8b3c07a841
    Update README.md Ross Wightman 2025-01-31 10:37:32 -0800
  • e025328f96
    simplify RNG Lucas Nestler 2025-01-31 17:26:14 +0100
  • 6367267298
    unify RNG Lucas Nestler 2025-01-31 17:23:53 +0100
  • 872978ccfe Fix comment, add 'stochastic weight decay' idea because why not Ross Wightman 2025-01-30 15:42:27 -0800
  • 510bbd5389 Change start/end args Ross Wightman 2025-01-30 13:13:49 -0800
  • 31831f5948 Change flattening behaviour in Kron Ross Wightman 2025-01-30 13:07:20 -0800
  • 5f85f8eefa Fix comment, add 'stochastic weight decay' idea because why not kron_flatten Ross Wightman 2025-01-30 15:42:27 -0800
  • 5940cc167f Change start/end args Ross Wightman 2025-01-30 13:13:49 -0800
  • 3be8b1abe4 Change flattening behaviour in Kron Ross Wightman 2025-01-30 13:07:20 -0800
  • cdbafd9057 Try to force numpy<2.0 for torch 1.13 tests, update newest tested torch to 2.5.1 Ross Wightman 2025-01-28 14:51:25 -0800
  • 93f44d1805 Try to force numpy<2.0 for torch 1.13 tests, update newest tested torch to 2.5.1 update_test_workflow Ross Wightman 2025-01-28 14:51:25 -0800
  • b1752eefb5 Fix missing model key in bulk validate results on error Ross Wightman 2025-01-28 13:20:40 -0800
  • b3a83b81d6 Prep Kron for merge, add detail to attributions note, README. Ross Wightman 2025-01-27 16:00:58 -0800
  • 67ef6f0a92 Move opt_einsum import back out of class __init__ Ross Wightman 2025-01-27 14:03:25 -0800
  • 9ab5464e4d More additions to Kron Ross Wightman 2025-01-27 13:09:09 -0800
  • 5f10450235 Some more kron work. Figured out why some tests fail, implemented a deterministic rng state load but too slow so skipping some tests for now. Ross Wightman 2025-01-26 20:44:09 -0800
  • cd21e80d03 Fiddling with Kron (PSGD) Ross Wightman 2025-01-24 14:16:21 -0800
  • 9c26c959eb Prep Kron for merge, add detail to attributions note, README. kron_optimizer Ross Wightman 2025-01-27 16:00:58 -0800
  • 80a0205725 Move opt_einsum import back out of class __init__ Ross Wightman 2025-01-27 14:03:25 -0800
  • 71d174180f More additions to Kron Ross Wightman 2025-01-27 13:09:09 -0800
  • f759d1213a Some more kron work. Figured out why some tests fail, implemented a deterministic rng state load but too slow so skipping some tests for now. Ross Wightman 2025-01-26 20:44:09 -0800
  • de2f5c6a05 Fiddling with Kron (PSGD) Ross Wightman 2025-01-24 14:16:21 -0800
  • b283c24f13 Update T2T_ViT Ryan 2025-01-24 00:05:22 +0800
  • bd46b550de Add T2T_ViT Ryan 2025-01-23 04:31:22 +0800
  • d81da93c16 Use import alias Adam J. Stewart 2025-01-22 10:08:18 +0100
  • 4de1abf837 timm: add __all__ to __init__ Adam J. Stewart 2025-01-11 14:31:41 +0100
  • 1e189be46d
    Use import alias Adam J. Stewart 2025-01-22 10:08:18 +0100
  • bda46f8e6f Add num_classes assertion after reset_classifier Ryan 2025-01-22 01:43:32 +0800
  • 17eabaad17 Fix RDNet forward call Ryan 2025-01-22 00:41:04 +0800
  • 80a4877376 Fix self.reset_classifier num_classes update Ryan 2025-01-22 00:40:16 +0800
  • 84631cb5c6 Add missing training flag to convert_sync_batchnorm Collin McCarthy 2025-01-21 11:41:48 -0800
  • 611775581a Add missing training flag to convert_sync_batchnorm Collin McCarthy 2025-01-21 11:41:48 -0800
  • a31bb8a879 Add num_classes assertion after reset_classifier Ryan 2025-01-22 01:43:32 +0800
  • 61bf9e1af1 Fix RDNet forward call Ryan 2025-01-22 00:41:04 +0800
  • c0b12fcb16 Fix self.reset_classifier num_classes update Ryan 2025-01-22 00:40:16 +0800
  • a8b90bf5e9 monkeypatch ml-decoder + metaformer Fredo Guan 2025-01-21 01:38:35 -0700