huggingface/accelerate

Pull Requests Commits

SunMarc committed 2 years ago

7f48cb52

SunMarc committed 2 years ago

e33aba73

SunMarc committed 2 years ago

068d5869

SunMarc committed 2 years ago

76043b40

SunMarc committed 2 years ago

31269920

SunMarc committed 2 years ago

656e15e4

SunMarc committed 2 years ago

1b21f9a6

SunMarc committed 2 years ago

f592aad8

SunMarc committed 2 years ago

b6923957

SunMarc committed 2 years ago

f973f0d5

only run big modeling test

SunMarc committed 2 years ago

56580b40

SunMarc committed 2 years ago

30ac26cf

Enable config for fsdp activation checkpointing (#2779)

helloworld1 committed 2 years ago

Verified befd87f0

Skip deepspeed test (#2776)

SunMarc committed 2 years ago

Verified abce3604

Fix small edge case in get_module_leaves (#2774)

SunMarc committed 2 years ago

Verified 27a607ea

fix minor typo (#2767)

TemryL committed 2 years ago

Verified aa21174d

optimize get_module_leaves speed (#2756)

BBuf committed 2 years ago

Verified 6cf1cc0a

Sets default to PyTorch defaults based on backend (#2758)

muellerzr committed 2 years ago

Verified bb465a9c

Enable sharded cpu resume (#2762)

muellerzr committed 2 years ago

Verified 67308ca6

Revert "Simplify CLI args validation and ensure CLI args take precedence over config file." (#2763)

muellerzr committed 2 years ago

Verified 63772f6a

fix cpu omp num threads set (#2755)

jiqing-feng committed 2 years ago

Verified 8798cf06

Fix sagemaker config (#2753)

muellerzr committed 2 years ago

Verified 47bb2dd5

Simplify CLI args validation and ensure CLI args take precedence over config file. (#2757)

Iain-S committed 2 years ago

Verified 724824ab

Fix duplicate environment variable check in multi-cpu condition (#2752)

yhna940 committed 2 years ago

Verified afc2c99e

Fix max_memory assignment (#2751)

SunMarc committed 2 years ago

Verified 0fb95a2d

LOMO / FIX: Support multiple optimizers (#2745)

younesbelkada committed 2 years ago

Verified 7ac153f4

Fix stacklevel in `logging` to log the actual user call site (instead of the call site inside the logger wrapper) of log functions (#2730)

luowyang committed 2 years ago

Verified 0f1b91bb

Fixed the problem of incorrect conditional judgment statement when configuring enable_cpu_affinity (#2748)

ji-huazhong committed 2 years ago

Verified d1eb44c8

Update modeling.py by adding try-catch section to skip the unavailable devices (#2681)

MeVeryHandsome committed 2 years ago

Verified 11a36328

Add feature to allow redirecting std streams into log files when using torchrun as the launcher. (#2740)

lyuwen committed 2 years ago

Verified 5cfe4094

Older