NielsRogge/transformers

Pull Requests Commits

Remove old Python code (#41226)

cyyever committed 76 days ago

Verified cf0887f6

:rotating_light: [`v5`] Remove headmasking (#41076)

vasqu committed 76 days ago

Verified 52f5eca7

[generate] cache missing custom generate file (#41216)

gante committed 76 days ago

Verified a80f05df

Align pull request template to bug report template (#41220)

tomaarsen committed 76 days ago

Verified 1f1e93e0

[ESM] add accepts_loss_kwargs=False to EsmPreTrainedModel (#41006)

pstjohn committed 76 days ago

Verified 2a596f5b

Trainer: Pass `num_items_in_batch` to `compute_loss` in `prediction_step` (#41183)

pramodith committed 76 days ago

Verified 3edd8048

Avoid assumption that model has config attribute in deepspeed (#41207)

tomaarsen committed 76 days ago

Verified 59035fd0

Wait for main process in _save_checkpoint to ensure best checkpoint exists (#40923)

ssharpe42 committed 76 days ago

Verified d9739778

Deprecate `half_precision_backend` (#41134)

SunMarc committed 76 days ago

Verified 06c04e08

Fix Qwen3-Omni audio_token_id serialization issue (#41192)

eun2ce committed 76 days ago

Verified 0e5a9756

docs/examples(speech): pin CTC commands to Hub datasets; add Windows notes (#41027)

tayo4christ committed 77 days ago

Verified 42c68251

Remove unnecessary Optional typing (#41198)

cyyever committed 77 days ago

Verified aaf1269d

[docs] Fix links (#41110)

stevhliu committed 77 days ago

Verified 4a02bc70

Embed interactive timeline in docs (#41015)

yonigozlan committed 77 days ago

Verified def4a37e

Fix docker quantization (#41201)

SunMarc committed 77 days ago

Verified 3e975acc

Fix 8bit bnb loading (#41200)

SunMarc committed 77 days ago

Verified 8635d8e7

Fix EXAONE-4.0 dummy id (#41089)

lkm2835 committed 77 days ago

Verified 1f0e9a47

Add EdgeTAM (#39800)

yonigozlan committed 77 days ago

Verified bd37c453

[`Kernels Attention`] Change fallback logic to error out on explicit kernels request and include FA3 (#41010)

vasqu committed 77 days ago

Verified c1db3868

Make quantizers good citizens loading-wise (#41138)

Cyrilvallez committed 77 days ago

Verified 5426edec

Separate docker images for Nvidia and AMD in benchmarking (#41119)

ahadnagy committed 77 days ago

Verified 399c589d

Fix attention sink implementation in flex attention (#41083)

SamuelBarryCS committed 77 days ago

Verified 52cbc7c8

fix(trainer): Avoid moving model with device_map (#41032)

The5cheduler committed 77 days ago

Verified de9a75f5

enable flex attention ut cases on XPU (#40989)

yao-matrix committed 77 days ago

Verified bcc0dae7

Bump hfh prerelease version (#41175)

Wauplin committed 77 days ago

Verified fcd483f0

Fix inaccurate train_tokens_per_second when resuming from checkpoint (#41156)

lilin-1 committed 77 days ago

Verified a3fa1d39

[v5] Remove `model_parallel` deprecated feature (#41166)

SunMarc committed 77 days ago

Verified ad74fba0

More typing fixes (#41102)

cyyever committed 77 days ago

Verified 38a08b6e

[tests] `CausalLMTester` automatically infers other test classes from `base_model_class` 🐛 🔫 (#41066)

gante committed 77 days ago

Verified 4fade114

[XPU] Add MXFP4 support for XPU (#41117)

YangKai0616 committed 77 days ago

Verified cdba28c3

Newer Older