huggingface/accelerate

Pull Requests Commits

Allow tensor parallelism next to replicated data parallelism

qgallouedec committed 33 days ago

32f72742

Update command export to use SCRIPT and SCRIPT_ARGS (#3906)

timonmerk committed 248 days ago

Verified b7f2212e

Fix KMP_AFFINITY incorrectly set for non-CPU training (#3912)

hexfaker committed 248 days ago

Verified c231e513

enhance new codes to XPU, and make them be device agnostic (#3890)

yao-matrix committed 248 days ago

Verified 00e4d264

fix bug when both cpu_ram_efficient_loading and cpu_offload are enabled (#3910)

kaixuanliu committed 248 days ago

Verified 363816a8

skip device mesh creation when deepspeed and sp_size >1 (#3914)

kashif committed 248 days ago

Verified 08e212a3

fix tp only bug (#3908)

sywangyi committed 261 days ago

Verified 38dadd95

SunMarc committed 262 days ago

36479b88

Remove 8bit force hook for bnb (#3907)

SunMarc committed 262 days ago

Verified 4f8db612

fix crash in optimizer.step when fsdp2 is enabled and model is bfloat16 (#3905)

sywangyi committed 262 days ago

Verified 758f3b18

bug: fsdp cannot load optimizor state using dcp (#3904)

flymin committed 263 days ago

Verified cdb2d1ff

SunMarc committed 266 days ago

Verified a89f2018

Np parall fix (#3900)

sywangyi committed 266 days ago

Verified 2d388f1b

change the default value of fsdp_min_num_params to int (#3902)

CodeMan62 committed 268 days ago

Verified c7a77460

Fix hf_device_map device comparison in prepare_model (#3895)

rezaqorbani committed 269 days ago

Verified 27ed776d

Upgrade GitHub Actions to latest versions (#3893)

salmanmkc committed 269 days ago

Verified 4bf9964f

Upgrade GitHub Actions for Node 24 compatibility (#3892)

salmanmkc committed 269 days ago

Verified a6266e01

Disable hook compile (#3888)

SunMarc committed 289 days ago

Verified db8835cc

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

shantanugupta2004 committed 290 days ago

Verified 16b6b3f7

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

pengdurice committed 290 days ago

Verified 11e97ab8

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

amanzoni1 committed 290 days ago

Verified 696e6cfe

Remove ipex (#3883)

yao-matrix committed 291 days ago

Verified 4826b66d

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

kaixuanliu committed 291 days ago

Verified 0348b70c

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

amanzoni1 committed 295 days ago

Verified fa6e13d0

add MS-AMP deprecation warnings (#3857)

neha222222 committed 298 days ago

Verified 471f7e42

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

FlorianVal committed 302 days ago

Verified b9ca0de6

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

shimizust committed 303 days ago

Verified 75983a59

Updating support of Megatron-LM (#3842)

pengdurice committed 303 days ago

Verified 7e384699

Fix TE internal imports (#3852)

ksivaman committed 305 days ago

Verified 65db8acc

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

kashif committed 308 days ago

Verified a72834ca

Older