huggingface/accelerate

Pull Requests Commits

Disable hook compile (#3888)

SunMarc committed 18 days ago

Verified db8835cc

Fix: Remove duplicate W&B initialization in offline mode (#3886)

shantanugupta2004 committed 19 days ago

Verified 16b6b3f7

Update support of Megatron-LM PR 2 (#3887)

pengdurice committed 19 days ago

Verified 11e97ab8

Fix KeyError in extract_model_from_parallel for partial torch.compile (#3881)

amanzoni1 committed 19 days ago

Verified 696e6cfe

Remove ipex (#3883)

yao-matrix committed 20 days ago

Verified 4826b66d

using `spawn` instead of `fork` for XPU device (#3884)

kaixuanliu committed 20 days ago

Verified 0348b70c

Fix FSDP2 tied embedding errors with targeted ValueError guidance (#3878)

amanzoni1 committed 24 days ago

Verified fa6e13d0

add MS-AMP deprecation warnings (#3857)

neha222222 committed 27 days ago

Verified 471f7e42

check if cuda is available before calling nvidia-smi (#3872)

FlorianVal committed 31 days ago

Verified b9ca0de6

Fix FP8 torchao default config with padding and FSDP2 all-gather support (#3831)

shimizust committed 32 days ago

Verified 75983a59

Updating support of Megatron-LM (#3842)

pengdurice committed 32 days ago

Verified 7e384699

Fix TE internal imports (#3852)

ksivaman committed 34 days ago

Verified 65db8acc

error out if both CP and SP enabled (#3862)

kashif committed 37 days ago

Verified a72834ca

[SP] fix loss computation example (#3858)

kashif committed 37 days ago

Verified b923f65a

fix module and optimizer parameter mismatch before prepare_tp_ (#3845)

naomili0924 committed 38 days ago

Verified ed1a3365

Allow non-Tensor values in a batch with `dispatch_batches=True` (#3850)

tomaarsen committed 39 days ago

Verified b521400f

Upcast FSDP2 parameters only if requires_grad (#3848)

ojh31 committed 39 days ago

Verified d1c96ba9

Update PR template, setup.py author email (#3851)

tomaarsen committed 40 days ago

Verified 0cca1ddb

feat: added fine tuning example focused on TPUs (#3847)

tengomucho committed 41 days ago

Verified 9a20895d

SunMarc committed 44 days ago

139d14b1

device type helper (#3843)

kashif committed 44 days ago

Verified e3655b93

Deepspeed Ulysses/ALST integration (#3817)

stas00 committed 45 days ago

Verified c7e59dd7

use self hosted runner (#3841)

SunMarc committed 45 days ago

Verified e8f241f8

[Bug] Update torch.optim.Optimizer parameter states after tensor parallelism (#3835)

naomili0924 committed 46 days ago

Verified 00b1b181

Fix typo in broadcast_object_list docstring (#3823)

wsntxxn committed 52 days ago

Verified a73fd3a1

Change arxiv to hf papers (#3834)

qgallouedec committed 55 days ago

Verified a08fb323

update bnb quantisation 4bit flag (#3828)

hbraith committed 61 days ago

Verified 61bcdaa4

Remove warning for `cpu_ram_efficient_loading` (#3816)

SunMarc committed 74 days ago

Verified a12beee3

SunMarc committed 76 days ago

0a0c9895

refactor: nit change for get_parameters_from_modules (code debt) (#3815)

kmehant committed 82 days ago

Verified 5998f862

Older