huggingface/transformers

Pull Requests Commits

update `kernels` to 0.4.3 (#37419)

ArthurZucker committed 248 days ago

Verified a2c2fb01

mark llama4 as not supported with fa2 (#37416)

winglian committed 248 days ago

Verified 0ddad2d6

Offloaded hybrid cache for Llama4 (#37401)

Cyrilvallez committed 248 days ago

Verified fbb2054e

Fix Llama4 offset (#37414)

Cyrilvallez committed 248 days ago

Verified 6d8b0b33

Restrict & Explain tp_plan for FBgemm (#37404)

MekkCyber committed 248 days ago

Verified f5865d32

Handle torch ver in flexattn (#37400)

Kh4L committed 248 days ago

Verified e39c7326

Add warning when failed to acquire other user's lock at model download (#37395)

manueldeprada committed 248 days ago

Verified bc0150bb

handle torch version edge cases (#37399)

winglian committed 248 days ago

Verified 9cda4265

the fix that did not get in (#37370)

ArthurZucker committed 248 days ago

Verified e032d12e

Attention Quantization with FBGemm & TP (#37384)

MekkCyber committed 248 days ago

Verified f834ca2c

Fix some failing AWQ tests (#37383)

DerekLiu35 committed 248 days ago

Verified c5c648dd

Apply torchfix to replace deprecated functions: `_pytree._register_pytree_node` and `torch.cpu.amp.autocast` (#37372)

bzhong-solink committed 248 days ago

Verified 71b35387

Fix warning message for PEFT models in text-generation pipeline #36783 (#36887)

falconlee236 committed 248 days ago

Verified ad340908

Add "selecting a quantization method" doc (#37159)

DerekLiu35 committed 249 days ago

Verified 2527f71a

update deepspeed docker (#37371)

SunMarc committed 249 days ago

Verified 7ae0be72

Add glm4 (#37388)

ArthurZucker committed 249 days ago

Verified e3eda6d1

fix: llama4 conversion script no_rope_layers (#37359)

jmkuebler committed 249 days ago

Verified 1e6ff5fd

Update composition flag usage (#36263)

zucchini-nlp committed 249 days ago

Verified 6f4058ae

Preserve requires_grad in pre quantized model (#37354)

jerryzh168 committed 249 days ago

Verified 08e3217b

:rotating_light: :rotating_light: Setup -> setupclass conversion (#37282)

Rocketknight1 committed 249 days ago

Verified 4d0de5f7

fix(qwen): fix shape error when using tp (#36947)

KimmiShi committed 249 days ago

Verified c15a7adb

prune LM Head for USD (#36695)

jmamou committed 249 days ago

Verified 121f91d3

[core] remove `GenerationMixin` inheritance by default in `PreTrainedModel` (#37173)

gante committed 249 days ago

Verified 4321b064

Skip non-selected experts for mixtral and qwen2_moe (#32429)

Coco58323 committed 249 days ago

Verified aab08783

[llama 4] dynamic rope decorator (#37365)

gante committed 249 days ago

Verified 35f0f5b5

Set vision config to None for Gemma 1B conversion (#37366)

RyanMullins committed 250 days ago

Verified 530322cc

fix deepspeed job (#37284)

ydshieh committed 250 days ago

Verified 8064cd9b

A bit of cleaning 🧹🧹 (#37215)

Cyrilvallez committed 250 days ago

Verified cdfb018d

Use Python 3.9 syntax in tests (#37343)

cyyever committed 250 days ago

Verified 1e6b546e

convert float for yarn related arguments in rope_scaling (#37139)

bzantium committed 250 days ago

Verified 0fc683d1

Newer Older