Commit Graph

Select branches

Hide Pull Requests

0.5.x

0.6.x

adafactor_bv

adamuon

aimv2

apple_mobileclip2

attention_pool_dim_handling

attn_update

augmentation_update

beit3_remap

beit3_remap_clean

bits_and_tpu

bulk_runner_tweaks

cache_dir

calculate_dpr

caojiaolong-main

cautious_optim

check_intermediate_feat

cleanup_fp32_norm

cleanup_vit_convert

cleanup_xla_model_fixes

clip_laion2b

convnext_and_copyright

convnext_nano_r384

convnext_shortcut

convnext_xxlarge_eps

convnext_zepto

convnextv2

convnormact_aa_none

coord_attn

corrected_weight_decay

csatv2-gusdlf93

dataset_info

dd_factory_kwargs

device_amp_cleanup

dfnvit

diff_parallel

diffattn

dinov2_reg4

dinov3

dist_loss_tracking

distill_fixups

dot_nine_cleanup

dtensor_muon

dynamic_resize_vit

edgenext_csp_and_more

efficientnet_attn

efficientnet_x

efficientnet_x_w

effnet_improve_features_only

endof2025_weights

eva

eva02

eva_input_size

eva_pe_integration

fcossio-vit-maxpool

feature_base

few_more_weights

fix_asymm_set_grad_enable

fix_drop_block

fix_hrnet_head

fix_imagenet22k_ms_mapping

fix_levit_load

fix_mnv4_query_strides

fix_onnx

fix_out_indices_order

fix_parallel_scaling_bias

fix_rope_xy

fix_swin_pad

fix_tests

fix_tests_210

fix_window_trace

fixes-syncbn_pretrain_cfg_resolve

fixes_bce_regnet

florence2

focalnet_and_swin_refactor

fp32_norm

freeze_norm_revisit

friday_weights

gemma3n_enc_weights

gh-pages

grodino-dataset_trust_remote

group_size_eff

half_prec_trainval

hf_hub_revisit

hparam_docs

improve_attn_mask

inception_next

inference_tests

init_non_persistent_buffers

intern300m

kron_flatten

kron_optimizer

legacy_adadamw_update

levit_efficientformer_redux

local_dir

main

mambaout

mars_tweak

maxxvit_hub

mega_merge

mesa_ema

meshgrid_ndgrid

metaformer_baselines_for_vision

misc_fixes_2026_02

mnv4_first_weights

mnv5_bias_str_norm

mobileclip_hub_id

mobilenetv5

mobilenetv5_final_uib_pad

mobileone

model_cleanup_docstrings_anno

more_attn

more_datasets

more_mobile

more_naflex

more_opt_stuff

more_tiny_test_models

more_vit

multi-weight

multi-weight_effnet_convnext

muon

muon_fallback

mw-deit

mw-final

mw-mlp_mixer

mw-more

mw-resnet

naflex

naflex_in1k_weights

naflex_resampling_speedups

naflex_rope

naflex_siglip2_weights

no_opt_layer_decay

non_reentrant_ckpt

norm_mlp_classifier

norm_norm_norm

onnx_dynamo

onnx_export

openai_clip_resnet

openclip_weight_move

opt_mars_more

other_pool

parse_kwargs

parse_kwargs2

patch_drop_refactor

perceiver

pos_embed_type_safe

preact_resnet18

prr_pool

pyproject_pdm

refactor-imports

refactor_pre_logits

remove_clamp_min

resnetv2_18_34

rope_vit

samvit_fix_and_rope

sbb2_vit_hiera_weights

scaling_vit

scheduler_types

script_cleanup

siglip2

siglip_update

sknet_kwargs_merge

small_384_weights

small_things

so150m2

ssl_tasks

stream_device

swin_maxvit_resize

swin_v2

tiny_test_models

token_distill_task

transforms_dataset_updates

update-test-pytorch

update_clip_pretrained

update_results_2025

update_test_workflow

vibe_rsvit

vit_and_bit_test_fixes

vit_packed

vit_relpos_refactor

vit_siglip_and_reg

warmup_prefix_fix

weights_only

weights_only_default_true

wip-voidbag-accumulate-grad

xsepconv

yehuitang-Add-GhostNetV2

yuweihao-inception_next_atto

#1

#1007

#101

#1012

#1014

#1034

#1037

#105

#1055

#1056

#1056

#1058

#1061

#1063

#1063

#1069

#1083

#1091

#1094

#1105

#1112

#1113

#115

#1150

#117

#1187

#1190

#1194

#12

#1208

#1210

#1212

#1213

#1213

#122

#1222

#1229

#123

#1230

#1233

#1239

#1245

#1249

#125

#1251

#1253

#1256

#1259

#1265

#1266

#1270

#1271

#1273

#1274

#1276

#1277

#1294

#13

#1317

#132

#1322

#1327

#1330

#1336

#1340

#1345

#1349

#1351

#1354

#136

#1363

#1365

#137

#1381

#1383

#1385

#140

#141

#1414

#1415

#142

#1420

#143

#1439

#145

#146

#1467

#1471

#1473

#1476

#1479

#148

#1483

#1493

#1494

#1498

#150

#1500

#1502

#1520

#1522

#1525

#1527

#1532

#1538

#154

#1545

#1547

#155

#1552

#1553

#156

#1562

#1564

#1565

#1575

#1578

#1581

#1582

#1583

#1586

#1590

#1592

#1593

#16

#1606

#1612

#1614

#1617

#1621

#1622

#1624

#1627

#1628

#163

#1630

#1633

#1641

#1643

#1645

#1647

#1649

#1654

#1655

#1659

#1662

#1673

#1674

#1677

#1679

#1680

#1683

#1685

#1687

#1688

#1690

#1691

#1696

#1708

#1714

#1715

#1719

#1720

#1720

#1736

#1741

#1745

#1746

#1747

#1748

#175

#1756

#1760

#1761

#1766

#1778

#1784

#1785

#1789

#179

#1793

#1797

#1799

#1804

#1812

#1823

#183

#1830

#1834

#1835

#1837

#1841

#1846

#1847

#1850

#1852

#1866

#1873

#1876

#1888

#1889

#1890

#1894

#19

#1900

#1903

#191

#1911

#1913

#1914

#1916

#1918

#1919

#1920

#1923

#1925

#1926

#1930

#1937

#1945

#195

#1951

#1952

#1954

#1955

#1958

#1961

#1964

#1965

#1967

#1971

#1973

#1974

#1975

#1978

#1980

#1985

#1988

#199

#1993

#1995

#1996

#1998

#2

#2001

#2002

#2003

#2005

#2008

#2010

#2011

#2012

#2014

#2015

#2016

#2017

#2019

#2020

#2023

#2029

#2032

#2032

#2033

#2034

#2036

#2037

#2039

#2040

#2041

#2043

#2045

#2045

#2048

#2049

#2050

#2051

#2052

#2053

#2058

#2062

#2062

#2064

#2071

#2077

#2079

#208

#2082

#2089

#2090

#2091

#2092

#2096

#2098

#2102

#2105

#2108

#2109

#2111

#2113

#2117

#212

#2121

#2123

#2124

#2126

#213

#2136

#2138

#214

#2142

#2143

#2144

#2145

#2156

#216

#2162

#2164

#2165

#2167

#2167

#2168

#2169

#2174

#2175

#2176

#2176

#2178

#218

#2180

#2181

#2182

#2183

#2189

#2191

#2193

#2195

#2196

#2197

#2198

#2199

#2200

#2202

#2203

#2205

#2207

#2209

#2213

#2214

#2217

#2220

#2223

#2225

#2229

#2230

#2236

#2238

#2239

#2240

#2243

#2245

#2246

#2252

#2253

#2257

#2258

#2263

#2264

#2266

#227

#2274

#2275

#2277

#228

#2281

#2285

#2292

#2297

#2298

#2299

#23

#2304

#2305

#2308

#2311

#2312

#2314

#2316

#2317

#2319

#2319

#2320

#2321

#2322

#2325

#2326

#2328

#233

#2331

#2333

#2334

#2336

#2337

#2340

#2341

#2343

#2346

#2347

#2349

#2350

#2350

#2351

#2353

#2356

#2357

#2358

#2359

#2360

#2361

#2369

#237

#2372

#2372

#2376

#2379

#2381

#2382

#2384

#2386

#2386

#2387

#2388

#2389

#2394

#2396

#2397

#2398

#2399

#24

#240

#2400

#2402

#2403

#2404

#2405

#2406

#2408

#2409

#2412

#2413

#2415

#2416

#2417

#2418

#2418

#2421

#2423

#2426

#2426

#2427

#2429

#2431

#2433

#2439

#244

#2440

#2441

#2442

#2444

#2448

#2466

#2467

#2471

#2474

#2475

#2476

#2477

#2478

#2480

#2482

#2483

#2484

#2486

#2487

#2489

#2490

#2490

#2496

#2499

#250

#2501

#2502

#2503

#2504

#2507

#2508

#2509

#2510

#2511

#2515

#2516

#2517

#2518

#2519

#2523

#2526

#2527

#2529

#2532

#2533

#2534

#2535

#2536

#2537

#2538

#2539

#2541

#2543

#2544

#2545

#255

#2552

#2553

#2554

#2559

#256

#2560

#2561

#2563

#2564

#2568

#2569

#2571

#2574

#2574

#2576

#2578

#2579

#2585

#2589

#2591

#2592

#2595

#2596

#2598

#2599

#2602

#2603

#2605

#2606

#2610

#2614

#2617

#2620

#2621

#2624

#2625

#2626

#2627

#2629

#263

#2630

#2631

#2632

#2633

#2634

#2635

#2636

#2637

#2638

#2639

#2640

#2641

#2642

#2643

#2645

#2647

#2648

#2649

#2650

#2650

#2654

#2655

#2656

#2656

#2657

#2658

#2662

#2664

#2665

#2665

#2666

#2668

#2669

#2672

#2673

#2674

#2675

#2676

#2677

#2677

#2678

#2679

#268

#2680

#2683

#2684

#2685

#2686

#282

#284

#286

#29

#290

#297

#30

#300

#302

#31

#315

#32

#322

#323

#33

#334

#335

#35

#352

#367

#382

#389

#401

#413

#418

#419

#421

#425

#427

#429

#434

#437

#440

#447

#450

#459

#475

#489

#494

#497

#501

#502

#510

#52

#525

#528

#53

#533

#537

#542

#548

#550

#554

#555

#556

#561

#571

#580

#581

#595

#6

#603

#609

#612

#613

#62

#625

#626

#636

#637

#645

#647

#65

#651

#652

#653

#659

#66

#660

#668

#669

#679

#681

#685

#693

#7

#702

#714

#729

#731

#738

#74

#746

#747

#750

#768

#771

#772

#800

#801

#804

#804

#805

#808

#812

#813

#816

#817

#82

#821

#828

#83

#841

#842

#853

#858

#87

#874

#875

#876

#88

#880

#882

#891

#898

#910

#913

#914

#915

#933

#94

#95

#964

#967

#968

#986

#989

#99

#993

v0.1-attn-weights

v0.1-cadene

v0.1-coat-weights

v0.1-dnf-weights

v0.1-effv2-weights

v0.1-ger-weights

v0.1-hrnet

v0.1-mvit-weights

v0.1-pit-weights

v0.1-regnet

v0.1-repvgg-weights

v0.1-res2net

v0.1-resnest

v0.1-rexnet

v0.1-rs-weights

v0.1-rsb-weights

v0.1-selecsls

v0.1-tpu-weights

v0.1-tresnet

v0.1-vitjx

v0.1-vt3p-weights

v0.1-weights

v0.1-weights-maxx

v0.1-weights-morevit

v0.1-weights-swinv2

v0.1.1

v0.2.1

v0.3.3

v0.3.4

v0.4.12

v0.4.5

v0.4.9

v0.5.4

v0.6.11

v0.6.12

v0.6.13

v0.6.5

v0.6.7

v0.8.10dev0

v0.8.13dev0

v0.8.17dev0

v0.8.2dev0

v0.8.6dev0

v0.9.0

v0.9.1

v0.9.10

v0.9.11

v0.9.12

v0.9.16

v0.9.2

v0.9.5

v0.9.6

v0.9.7

v0.9.8

v0.9.9

v1.0.10

v1.0.11

v1.0.12

v1.0.13

v1.0.14

v1.0.15

v1.0.16

v1.0.17

v1.0.18

v1.0.19

v1.0.20

v1.0.21

v1.0.22

v1.0.23

v1.0.24

v1.0.25

v1.0.26

v1.0.3

v1.0.7

v1.0.8

v1.0.9

Mono Color

• 8d0f79effa Release 1.0.26 main v1.0.26 Ross Wightman 2026-03-23 11:09:21 -07:00
• 6e3fdda395 Implement PRR as a pooling module. Alternative to #2678 Ross Wightman 2026-03-13 13:38:35 -07:00
• 8b4239c4d5 Add comments for DinoV3 re global pool (class token). Fix #2681 Ross Wightman 2026-03-17 09:02:59 -07:00
• 52e6d19d9d Change avg_checkpoints.py to use more secure load helper Ross Wightman 2026-03-16 14:06:34 -07:00
• 7a2f49bd49 Fix FX tracing on resolve_self_attn_mask Ross Wightman 2026-03-16 11:48:53 -07:00
• 61a26c7707 Improve attention mask handling for vision_transformer and eva and related blocks Ross Wightman 2026-03-16 10:42:35 -07:00
• 5681186444 Fix FX tracing on resolve_self_attn_mask improve_attn_mask Ross Wightman 2026-03-16 11:48:53 -07:00
• f59f8e0a3a Improve attention mask handling for vision_transformer and eva and related blocks Ross Wightman 2026-03-16 10:42:35 -07:00
• 682c8458d0 Implement PRR as a pooling module. Alternative to #2678 prr_pool Ross Wightman 2026-03-13 13:38:35 -07:00
• 3e8def86c4 Improve 2d and latent attention pool dimension handling. Fix #2682 Ross Wightman 2026-03-13 10:45:36 -07:00
• f8c695d164 Improve 2d and latent attention pool dimension handling. Fix #2682 attention_pool_dim_handling Ross Wightman 2026-03-13 10:45:36 -07:00
• a94c10fce1 Update version.py Ross Wightman 2026-03-10 07:40:16 -07:00
• 0c90043d23 fix: branch Hiera MaskUnitAttention into 4D global path for FlashAttention dispatch Your Name 2026-03-09 17:45:02 +05:30
• a346c76b5f Further refine weights_only=True, add safe globals for argparse Namespace to avoid failures on timm train checkpoints Ross Wightman 2026-03-09 11:11:17 -07:00
• 7b439f4ba4 default weights_only=True for load fns Ross Wightman 2026-03-08 21:44:12 -07:00
• a13b9b84ec Further refine weights_only=True, add safe globals for argparse Namespace to avoid failures on timm train checkpoints weights_only_default_true Ross Wightman 2026-03-09 11:11:17 -07:00
• 2f750c0962 default weights_only=True for load fns Ross Wightman 2026-03-08 21:44:12 -07:00
• fa0d31e88c Fix CLS and Reg tokens usage when pos_embed is disabled Sina Hajimiri 2026-03-05 14:42:09 -05:00
• 85bb3304d2 Enhance SGDP optimizer with caution parameter (#2675) Jinghui Yuan 2026-03-06 02:48:07 +08:00
• f4f4cfc73f Add timmx model export tool to README Youssef Boulaoaune 2026-02-26 10:39:56 +09:00
• 2845814ea6 fix: replace bare except clauses with except Exception haosenwang1018 2026-02-25 14:20:02 +00:00
• 9326ff27f0 Release 1.0.25 v1.0.25 Ross Wightman 2026-02-23 08:45:52 -08:00
• 75684faa04 Add DTensor compatible NS impl for Muon Ross Wightman 2026-02-18 09:26:59 -08:00
• 4b727ccb70 Add DTensor compatible NS impl for Muon dtensor_muon Ross Wightman 2026-02-18 09:26:59 -08:00
• feed46c9ac Change clamp_min_ to clamp_(min=) as former doesn't work with DTensor / FSDP2 without custom registration Ross Wightman 2026-02-15 11:02:50 -08:00
• 94b799bde7 Change clamp_min_ to clamp_(min=) as former doesn't work with DTensor / FSDP2 without custom registration remove_clamp_min Ross Wightman 2026-02-15 11:02:50 -08:00
• 56467ad5f2 fix(optim): replace bare except with Exception in Lion optimizer Luka Aladashvili 2026-02-11 13:20:04 +04:00
• 2490051330 Add a StepGate/StepEmbedding idea to RSViT variants. Add stochastic supervision/recursion loop options. vibe_rsvit Ross Wightman 2026-02-10 13:36:33 -08:00
• 6f1005af19 Committing what's there. Still a big WIP, in the middle of iterating on config layout and arg mappings (not ideal yet). ssl_tasks Ross Wightman 2026-02-06 12:44:54 -08:00
• 41754f29d3 Fix #2661 ... don't skip reset_parameters/init when meta device detected as it breaks use of accelerate and similar dispatch override context managers Ross Wightman 2026-02-06 10:44:40 -08:00
• bdab30b0eb Remove torch.jit calls except for .ignore and .is_scripting to avoid 2.11 deprecation. Fix #2663 Ross Wightman 2026-02-06 10:29:37 -08:00
• 31d987afb6 Fix #2661 ... don't skip reset_parameters/init when meta device detected as it breaks use of accelerate and similar dispatch override context managers misc_fixes_2026_02 Ross Wightman 2026-02-06 10:44:40 -08:00
• 7a08ab54fa Remove torch.jit calls except for .ignore and .is_scripting to avoid 2.11 deprecation. Fix #2663 Ross Wightman 2026-02-06 10:29:37 -08:00
• e8f0ae34b4 Fixup LayerScale use in perceiver & rsvit Ross Wightman 2026-01-30 10:30:56 -08:00
• 9171d82efc Enhance the numerical stability of the Cautious Optimizer Jinghui Yuan 2026-01-29 14:51:58 +08:00
• 17b976408d Add the test of cadamp Jinghui Yuan 2026-01-28 13:21:51 +08:00
• ba85080efd Add cadamp optimizer to optimization factory Jinghui Yuan 2026-01-28 13:03:44 +08:00
• a3643d20d2 Adding the cautious optimizer and the spherical cautious optimizer Jinghui Yuan 2026-01-28 12:59:41 +08:00
• 3cbb8f91c0 Fix perceiver tests Ross Wightman 2026-01-28 09:46:27 -08:00
• 125cd184fc Fix use with soft target loss, and tests for FX / classifier Ross Wightman 2026-01-28 08:34:33 -08:00
• cfb77f965e Vibe coding some different ideas on vit param sharing/recursion, latent/perceiver structure. Ross Wightman 2026-01-27 10:45:46 -08:00
• c9e22eb61e More task work, add engine, initial timm.apps ssl/cls scripts, debugging nepa and lejepa training Ross Wightman 2026-01-16 20:25:33 -08:00
• 9800e14d9c POC work on ssl (NEPA/JEPA/AIM variants), expansion of task framework Ross Wightman 2026-01-14 14:09:50 -08:00
• 370df82133 Update README.md Ross Wightman 2026-01-21 14:32:44 -08:00
• 247c545fa1 Fix #2653, no models with weights impacted so just a clean fix, remove buffer as benchmarks w/ pt 2.9 show no difference Ross Wightman 2026-01-20 15:40:27 -08:00
• 0d6b2cb8a5 Update README.md fix_parallel_scaling_bias Ross Wightman 2026-01-21 14:32:44 -08:00
• 824e98ff48 Fix #2653, no models with weights impacted so just a clean fix, remove buffer as benchmarks w/ pt 2.9 show no difference Ross Wightman 2026-01-20 15:40:27 -08:00
• 836dd99075 Fix distilled head dropout using wrong token in PiT forward_head Ofer Hasson 2026-01-18 18:12:57 +02:00
• 3c6f79df77 Remove mistaken .in_chans attr set in classifier class Ross Wightman 2026-01-09 11:26:06 -08:00
• 7945d3a014 Refactor distillation tasks to allow creation of DistillationTecher within Task. Add TokenDistillationTask. Ross Wightman 2026-01-09 10:49:34 -08:00
• c0e3d24fe6 Add .in_chans as an attribute to all models Ross Wightman 2026-01-09 10:06:27 -08:00
• f87dfb9b0d Dev version 1.0.25 Ross Wightman 2026-01-09 13:24:25 -08:00
• 628f47936b Remove mistaken .in_chans attr set in classifier class token_distill_task Ross Wightman 2026-01-09 11:26:06 -08:00
• 1d2b50c7a4 Refactor distillation tasks to allow creation of DistillationTecher within Task. Add TokenDistillationTask. Ross Wightman 2026-01-09 10:49:34 -08:00
• f6673802ea Add .in_chans as an attribute to all models Ross Wightman 2026-01-09 10:06:27 -08:00
• 90cae8c5ab Add new inference benchmark results v1.0.24 Ross Wightman 2026-01-06 15:23:54 -08:00
• ed276693a2 Remove ViT B/16 384 augreg2 tag that doesn't exist Ross Wightman 2026-01-06 15:21:25 -08:00
• 48625b172d Update README.md with release / fix info. Version 1.0.24 Ross Wightman 2026-01-06 09:17:18 -08:00
• 030985f285 Fix #2644 with full import path Ross Wightman 2026-01-06 07:35:40 -08:00
• b6432177e3 Version 1.0.23 v1.0.23 Ross Wightman 2026-01-05 13:39:54 -08:00
• 31a85de79a Update typing in other scheduler classes. Fix spacing in cosine typing. Add some basic scheduler unit tests Ross Wightman 2025-12-31 11:15:20 -08:00
• ec0e81edcc Update typing in other scheduler classes. Fix spacing in cosine typing. Add some basic scheduler unit tests scheduler_types Ross Wightman 2025-12-31 11:15:20 -08:00
• 8038db635d Update README.md Ross Wightman 2025-12-30 15:57:13 -08:00
• e33cd718a5 Add csatv2_21m weights at 512 & 640 img size. Add layer-scale support to csatv2 but not used yet. Ross Wightman 2025-12-29 16:23:57 -08:00
• 816a45f02e Add sbb vit 'dlittle' weights trained with NAdaMuon Ross Wightman 2025-12-28 14:09:44 -08:00
• 9d62118f3d refactor(scheduler): add type hints to CosineLRScheduler Yohei Okabayashi 2025-12-30 18:05:19 +09:00
• 8674fa41ea Add csatv2_21m weights at 512 & 640 img size. Add layer-scale support to csatv2 but not used yet. endof2025_weights Ross Wightman 2025-12-29 16:23:57 -08:00
• 940edf4285 Add sbb vit 'dlittle' weights trained with NAdaMuon Ross Wightman 2025-12-28 14:09:44 -08:00
• d4ab5165b4 Add docstrings to layer helper functions and modules (#2634) Murat Raimbekov 2025-12-26 22:17:26 +06:00
• 3eb784a6dc Rejig init_weights for 'skip' mode, add 'reset' to make it work with meta init Ross Wightman 2025-12-24 10:20:45 -08:00
• 741d9997ac Fix efficientvit msra check Ross Wightman 2025-12-23 07:19:39 -08:00
• 6bb1efc49c Use iterator for vit is_meta check, too configurable Ross Wightman 2025-12-22 21:46:14 -08:00
• 50993861cb Add 'skip' to vit init_weights assert with new sequencing Ross Wightman 2025-12-22 17:47:52 -08:00
• 2902c0cf29 Use blocks for is_meta check as patch_embed changes Ross Wightman 2025-12-22 15:51:06 -08:00
• b5ffe5bd99 Further refine non-persistent buffer init and ensure reasonable overlap with weight init, weight init moving in direction of planned changes for overlapping models Ross Wightman 2025-12-22 15:11:14 -08:00
• d2553fcdff Initial pass of an 'init_non_persistent_buffers' scheme, WIP.. needs more test and probably missed a few things Ross Wightman 2025-12-19 14:09:10 -08:00
• fdf3b42fc3 Rejig init_weights for 'skip' mode, add 'reset' to make it work with meta init init_non_persistent_buffers Ross Wightman 2025-12-24 10:20:45 -08:00
• c59c9c1a29 Modify autocasting in fast normalization functions to handle optional weight parameters safely. Mattie Tesfaldet 2025-12-19 13:12:09 -05:00
• d914bef04c Fix efficientvit msra check Ross Wightman 2025-12-23 07:19:39 -08:00
• eb64c3baad Use iterator for vit is_meta check, too configurable Ross Wightman 2025-12-22 21:46:14 -08:00
• b1e5167f68 Add 'skip' to vit init_weights assert with new sequencing Ross Wightman 2025-12-22 17:47:52 -08:00
• d144ff1071 Use blocks for is_meta check as patch_embed changes Ross Wightman 2025-12-22 15:51:06 -08:00
• 782c8d9807 Further refine non-persistent buffer init and ensure reasonable overlap with weight init, weight init moving in direction of planned changes for overlapping models Ross Wightman 2025-12-22 15:11:14 -08:00
• 431c1b66b8 Upgrade GitHub Actions for Node 24 compatibility Salman Muin Kayser Chishti 2025-12-20 23:40:37 +00:00
• f12ac6aa11 Initial pass of an 'init_non_persistent_buffers' scheme, WIP.. needs more test and probably missed a few things Ross Wightman 2025-12-19 14:09:10 -08:00
• 4e651dada1 Add HParams sections to hfdocs (#2630) Ross Wightman 2025-12-16 10:33:54 -08:00
• c1425cad03 Update hparams.mdx hparam_docs Ross Wightman 2025-12-16 10:25:42 -08:00
• a0fd395c17 Add HParams sections to hfdocs Ross Wightman 2025-12-16 10:19:19 -08:00
• de61fce2a5 Update README.md Ross Wightman 2025-12-16 08:53:34 -08:00
• ff95890e6a Update changes.mdx Ross Wightman 2025-12-16 08:45:31 -08:00
• 962559deeb Update changes.mdx Ross Wightman 2025-12-16 08:44:15 -08:00
• d7cfda4c78 Update README.md Ross Wightman 2025-12-16 08:43:19 -08:00
• 86e9d21bc5 Add nadamuon trained dwee / dpwee vit weights. Add comment to muon impl Ross Wightman 2025-12-15 10:19:52 -08:00
• c263de1872 Update README.md Ross Wightman 2025-12-12 11:34:50 -08:00
• 7f8fd8a30d Csatv2 contribution (#2627) Ross Wightman 2025-12-12 11:21:42 -08:00
• 81ec5a1e7d Unfold not needed with stride=kernel_size, removed unecessary permutes for a speedup csatv2-gusdlf93 Ross Wightman 2025-12-12 10:40:43 -08:00
• dd57311c53 DCT stats list -> tuple Ross Wightman 2025-12-12 10:21:40 -08:00
• 7a75178134 Upload csatv2 weights to hub, fix non-contiguous dct weight Ross Wightman 2025-12-12 10:09:34 -08:00
• 24fd7d2f32 Compact checkpiont_filter_fn a bit, make learnable dct out dim changeable and define another model class to have another arch config tested. Ross Wightman 2025-12-12 09:45:59 -08:00
• b6eb61aba5 Another round of consistency changes for csatv2, make stage building dynamic for other network shapes, allow drop path option for transformer blocks. Ross Wightman 2025-12-11 15:13:01 -08:00