huggingface/accelerate

Pull Requests Commits

muellerzr committed 1 year ago

cac3c554

muellerzr committed 1 year ago

45d8e9b6

Echo date in env

muellerzr committed 1 year ago

c325e8ee

Dont push until date is right

muellerzr committed 1 year ago

8dc19494

muellerzr committed 1 year ago

bb1dce14

muellerzr committed 1 year ago

1fd14139

Test with CPU runner

muellerzr committed 1 year ago

0e12b080

muellerzr committed 1 year ago

f0e7acad

Try with build to raw accelerate

muellerzr committed 1 year ago

8acc6d24

muellerzr committed 1 year ago

bb141ecd

XLA docker build

muellerzr committed 1 year ago

64d99d63

Make xla docker image

muellerzr committed 1 year ago

305f1055

muellerzr committed 1 year ago

ee163b66

Use uv instead of pip install for github CI (#2546)

muellerzr committed 1 year ago

Verified 354db5b5

Update FSDP mixed precision setter to enable fsdp+qlora (#2544)

pacman100 committed 1 year ago

Verified 92b1ad01

Allow Gradients to be Synced Each Data Batch While Performing Gradient Accumulation (#2531)

fabianlim committed 1 year ago

Verified 60bfdaa9

Remove extra double-dash in error message (#2541)

drscotthawley committed 1 year ago

Verified 16eb6d76

[docs] Troubleshoot (#2538)

stevhliu committed 1 year ago

Verified c8acfa70

Overdue email change... (#2534)

muellerzr committed 1 year ago

Verified e70e3c87

vmoens committed 1 year ago

Verified bc8dfe3c

Check if the buffers fit GPU memory after device map auto inferred (#2412)

notsyncing committed 1 year ago

Verified e3d32424

Update link to dynamo/compile doc (#2533)

thethomasboyer committed 1 year ago

Verified 10882eed

Update the default behavior of `zero_grad(set_to_none=None)` (#2472)

yongchanghao committed 1 year ago

Verified 145a98fc

Enable using dash or underscore for CLI args (#2527)

muellerzr committed 1 year ago

Verified 64ae9ea3

Launch mpirun from accelerate launch for multi-CPU training (#2493)

dmsuehir committed 1 year ago

Verified 8aa72b97

Remove unnecessary `env=os.environ.copy()`s (#2449)

akx committed 1 year ago

Verified 97d115a2

qbitstensor compatibility (#2526)

SunMarc committed 1 year ago

Verified 63cfd9ef

Don't manage `PYTORCH_NVML_BASED_CUDA_CHECK` when calling `accelerate.utils.imports.is_cuda_available()` (#2524)

luiscape committed 1 year ago

Verified 6cf8221a

Add copyright + some ruff lint things (#2523)

muellerzr committed 1 year ago

Verified 7a2feeca

fix typo in launch.py (#2516)

DerrickWang005 committed 1 year ago

Verified ee004674

Older