huggingface/transformers

Pull Requests Commits

Fix the FA2 logic in the longcat_flash model (#42549)

YangKai0616 committed 88 days ago

Verified c0328af6

Fix mixed torch.Tensor and DTensor in generate when use FSDP2 + LoRA (#42436)

Xiao-Chenguang committed 88 days ago

Verified 17c7c496

fix : cast into floats AFTER all assignments (#42587)

marconaguib committed 88 days ago

Verified 130dc471

fix: correct typos in code comments (#42577)

arrdel committed 88 days ago

Verified 8ba286ba

fix the FSDP1 default for reshard_after_forward (#42578)

winglian committed 88 days ago

Verified cba7ae86

Fix loaded data order bug when resuming from epoch >= 1 (#40691)

ngazagna-qc committed 88 days ago

Verified 629c0da4

[XPU] Fix fp8 UT patch func (#42584)

YangKai0616 committed 88 days ago

Verified 75c135d7

Use `getattr` in `standardize_rope_params` because `rope_parameters` not always present (#42593)

hmellor committed 88 days ago

Verified 232ecf2c

[Quantization] fix fbgemm (#42561)

MekkCyber committed 88 days ago

Verified 15b79ea8

make sure the FSDP plugin args are appropriately cast to bools (#42566)

winglian committed 88 days ago

Verified 377a8ee7

[Quantization] per tensor quantization kernel (#42560)

MekkCyber committed 88 days ago

Verified 51c5a7a6

[`CI`] Fix copies (#42571)

vasqu committed 88 days ago

Verified a5c061d2

Allow fallback to loading from Auto"SubProcessor".from_pretrained when model_type can't be inferred from config (#42402)

yonigozlan committed 88 days ago

Verified 675e8763

[SAM3] Compute masks once instead of per-layer, fix fa2 crash (#42543)

yonigozlan committed 88 days ago

Verified 80b408d1

Transformers serve fix (#42570)

LysandreJik committed 88 days ago

Verified 52b988d8

Kernel mapping error resolve (#42466)

Aaraviitkgp committed 88 days ago

Verified 3f174109

fix regression (#42569)

SunMarc committed 88 days ago

Verified 53d2bf6d

Fix `parse_response` after tokenizer refactor (#42300)

Rocketknight1 committed 88 days ago

Verified 5690f24e

Add backward compatibility for methods which have been moved to `RotaryEmbeddingConfigMixin` (#42517)

hmellor committed 89 days ago

Verified 0ba8f001

Fix eetq quanto quant methods (#42557)

SunMarc committed 89 days ago

Verified 5efd0d4a

[test] delete `SeamlessM4TProcessorTest::test_save_load_pretrained_addition` (#42562)

ydshieh committed 89 days ago

Verified 4ec83fe9

Fix fp8 + some enhancement (#42455)

SunMarc committed 89 days ago

Verified bc7a268f

[Fix] dots1 expert bias routing (#41663)

fjosw committed 89 days ago

Verified 29e8522b

[Ministral 3] Small fix config (#42537)

patrickvonplaten committed 89 days ago

Verified 64d8cf4f

[Quantization] fix dequant when block size is none & static quantization (#42545)

MekkCyber committed 89 days ago

Verified bb09a30f

Fix failing test in Glm4vMoeIntegrationTest (#42488)

Sai-Suraj-27 committed 89 days ago

Verified ac0769cd

Fix Qwen-VL family with prompt tuning (#42508)

zucchini-nlp committed 89 days ago

Verified 57eeb9ca

Add FastVLM (#41112)

camilla-deckard committed 89 days ago

Verified a6497675

Fix ernie moe (#42535)

vasqu committed 89 days ago

Verified 0fa49db1

[Ministral 3] Add ministral 3 (#42498)

patrickvonplaten committed 89 days ago

Verified bf3f0ae7

Newer Older