huggingface/accelerate

Pull Requests Commits

Rm clean_state_dict

muellerzr committed 2 years ago

170780d9

device agnostic cli/data_loader/grad_sync/kwargs_handlers/memory_utils testing (#2356)

wangshuai09 committed 2 years ago

Verified f88661b5

Add `adapter_only` option to `save_fsdp_model` and `load_fsdp_model` to only save/load PEFT weights (#2321)

AjayP13 committed 2 years ago

Verified 581fabba

modified big_modeling.py (#2376)

BlackSamorez committed 2 years ago

Verified e909eb34

`add_hook_to_module` and `remove_hook_from_module` compatibility with fx.GraphModule (#2369)

fxmarty committed 2 years ago

Verified 7644a02e

device agnosic optimizer testing (#2363)

ji-huazhong committed 2 years ago

Verified 162a8216

remove init_hook_kwargs (#2365)

fxmarty committed 2 years ago

Verified 0d6a5fa8

Fix deepspeed issue (#2366)

muellerzr committed 2 years ago

Verified 53845d25

bugfix that doesnt let fp8recipekwarg use TE or MSAMP (#2355)

sudhakarsingh27 committed 2 years ago

Verified 5ec00da2

fix test (#2354)

fxmarty committed 2 years ago

Verified 649e65b5

Fix `block_size` picking in megatron_lm_gpt_pretraining.py (#2342)

nilq committed 2 years ago

Verified 14d7c3fc

Fix mpi4py/failing deepspeed test issues (#2353)

muellerzr committed 2 years ago

Verified c7d11d7e

device agnostic test_accelerator/test_multigpu (#2343)

wangshuai09 committed 2 years ago

Verified ec4f01a0

FIX: add oneCCL environment variable for non-MPI launcher (accelerate launch) (#2339)

faaany committed 2 years ago

Verified f5c01eeb

Show DeepSpeed option when multi-XPU is selected in `accelerate config` (#2346)

faaany committed 2 years ago

Verified 20ff458d

Avoid duplicating memory for tied weights in `dispatch_model`, and in forward with offloading (#2330)

fxmarty committed 2 years ago

Verified 6719cb6d

Just 40* (#2332)

muellerzr committed 2 years ago

Verified 31fd2b1a

Fixed typos in readme files of docs folder. (#2329)

rishit5 committed 2 years ago

Verified fce61a99

Fix model memory issue (#2327)

muellerzr committed 2 years ago

Verified 6ec92cf0

convert it back to dict (#2326)

SunMarc committed 2 years ago

Verified 2a403732

Raise error when using batches of different sizes with `dispatch_batches=True` (#2325)

SunMarc committed 2 years ago

Verified f823404f

Update versions to dev

muellerzr committed 2 years ago

ef2fe912

Fix infer_auto_device_map when tied weights share the same prefix name (#2324)

fxmarty committed 2 years ago

Verified e3e9b875

`Params4bit` added to bnb classes in set_module_tensor_to_device() (#2315)

poedator committed 2 years ago

Verified 456afd92

fix sanity check (#2310)

izhx committed 2 years ago

Verified 0d2280da

Bring old seed technique back (#2319)

muellerzr committed 2 years ago

Verified 55d4a496

Update test_deepspeed.py (#2323)

pacman100 committed 2 years ago

Verified 2a8829d9

Fix DeepSpeed related regression (#2304)

pacman100 committed 2 years ago

Verified 3969731c

DeepSpeed refactoring (#2313)

pacman100 committed 2 years ago

Verified 411aa58a

Update accelerator.py (#2295)

stas00 committed 2 years ago

Verified 4420ec64

Older