huggingface/transformers

Pull Requests Commits

fix attention mask for flash attention

qgallouedec committed 274 days ago

bfb954fa

fix unable to disable compile

qgallouedec committed 274 days ago

e34b0733

Recover 2D mask when 4D is passed

qgallouedec committed 274 days ago

9e52db88

Fix post_init() code duplication (#36727)

Cyrilvallez committed 274 days ago

Verified 2c2495cc

🌐 [i18n-KO] Translated codegen.md to Korean (#36698)

maximizemaxwell committed 274 days ago

Verified 25992b49

[tests] Parameterized `test_eager_matches_sdpa_inference` (#36650)

gante committed 275 days ago

Verified 42ebb6c2

Try working around the processor registration bugs (#36184)

Rocketknight1 committed 275 days ago

Verified 9215cc62

Fix/best model checkpoint fix (#35885)

seanswyi committed 275 days ago

Verified 691d1b52

[model loading] don't `gc.collect()` if only 1 shard is used (#36721)

gante committed 275 days ago

Verified 3bd1a0dd

Cleanup the regex used for doc preprocessing (#36648)

Rocketknight1 committed 275 days ago

Verified 8cb522b4

Make the flaky list a little more general (#36704)

Rocketknight1 committed 275 days ago

Verified 72861e11

Gemma3 processor typo (#36710)

Kuangdd01 committed 275 days ago

Verified 53742b11

Add support for fast image processors in add-new-model-like CLI (#36313)

yonigozlan committed 275 days ago

Verified 69bc8484

Final CI cleanup (#36703)

Rocketknight1 committed 275 days ago

Verified 48ef468c

Add GGUF support to T5-Encoder (#36700)

Isotr0py committed 275 days ago

Verified b070025a

Handling an exception related to HQQ quantization in modeling (#36702)

MekkCyber committed 275 days ago

Verified 4a60bae8

fix: fsdp sharded state dict wont work for save_only_model knob (#36627)

kmehant committed 276 days ago

Verified 09a309d2

Add loading speed test (#36671)

Cyrilvallez committed 276 days ago

Verified 2a004f9f

[CI] Automatic rerun of certain test failures (#36694)

gante committed 276 days ago

Verified a3201cea

chore: fix typos in utils module (#36668)

threewebcode committed 276 days ago

Verified d8456938

Fix dtype for params without tp_plan (#36681)

Cyrilvallez committed 276 days ago

Verified 32c95bd8

fix type annotation for ALL_ATTENTION_FUNCTIONS (#36690)

WineChord committed 276 days ago

Verified bb965d8e

Change Qwen2_VL image processors to have init and call accept the same kwargs (#36207)

yonigozlan committed 276 days ago

Verified 1c287aec

Upgrading torch version and cuda version in quantization docker (#36264)

MekkCyber committed 276 days ago

Verified 65b8e38a

fix wandb hp search unable to resume from sweep_id (#35883)

bd793fcb committed 276 days ago

Verified 87b30c35

Changing the test model in Quanto kv cache (#36670)

MekkCyber committed 276 days ago

Verified 47cc4da3

Fix slicing for 0-dim param (#36580)

SunMarc committed 276 days ago

Verified bc3d5781

Update config.torch_dtype correctly (#36679)

SunMarc committed 276 days ago

Verified fbb18ce6

[Cache] Don't initialize the cache on `meta` device (#36543)

gante committed 276 days ago

Verified c4161238

Fix rescale normalize inconsistencies in fast image processors (#36388)

yonigozlan committed 276 days ago

Verified 79254c9b

Older