huggingface/accelerate

Pull Requests Commits

Release: v0.24.0

muellerzr committed 2 years ago

00301b27

muellerzr committed 2 years ago

Verified eb8c535c

Warn when kernel version is too low on Linux (#2077)

BenjaminBossan committed 2 years ago

Verified b7686ccb

fix docstring typo (#2072)

qgallouedec committed 2 years ago

Verified f3229872

Allow for samplers to be seedable and reproducable (#2057)

muellerzr committed 2 years ago

Verified 7843286f

Let iterable dataset shard have a len (#2066)

muellerzr committed 2 years ago

Verified 11e2e99c

DOC: Fix broken link to designing a device map (#2073)

BenjaminBossan committed 2 years ago

Verified 07e745f1

fix: remove useless token (#2069)

rtrompier committed 2 years ago

Verified c7c99a30

remove unused constants (#2045)

ji-huazhong committed 2 years ago

Verified 8f45a2ea

Fix the error when the "train_batch_size" is absent in DeepSpeed config (#2060)

LZHgrla committed 2 years ago

Verified 9fd64b7e

Add space to docs (#2055)

muellerzr committed 2 years ago

Verified 5be16ad9

Reset state to pass failing test

muellerzr committed 2 years ago

dab62832

Fix stalebot (#2052)

muellerzr committed 2 years ago

Verified caa9f9bc

fix docstring (#2053)

zhangsibo1129 committed 2 years ago

Verified 943efedb

Let drop_last modify `gather_for_metrics` (#2048)

muellerzr committed 2 years ago

Verified 50acb0c2

Make fsdp ram efficient loading optional (#2037)

pacman100 committed 2 years ago

Verified e6d96e5f

Fix integration CI (#2047)

muellerzr committed 2 years ago

Verified 1dfb6e93

Safely end training even if trackers weren't initialized (#1994)

Ben-Epstein committed 2 years ago

Verified 4bef6bc5

Reduce memory by using `all_gather_into_tensor` (#1968)

muellerzr committed 2 years ago

Verified 73640d04

Unpin transformers (#2044)

muellerzr committed 2 years ago

Verified 7a115914

Fix DeepSpeed version to <0.11 (#2043)

BenjaminBossan committed 2 years ago

Verified cbb0b82f

Allow FSDP to use with `torch.autocast` for bfloat16 mixed precision (#2033)

brcps12 committed 2 years ago

Verified 5ae61111

Fix save on each node (#2036)

muellerzr committed 2 years ago

Verified 230a5f54

Enable shared file system with `save` and `save_state` via ProjectConfiguration (#1953)

muellerzr committed 2 years ago

Verified 956114ac

update fsdp docs (#2026)

pacman100 committed 2 years ago

Verified 76ee7f21

Sync states for xpu fsdp (#2005)

abhilash1910 committed 2 years ago

Verified 420743af

update torch_dynamo backends (#1992)

SunMarc committed 2 years ago

Verified 206ab491

Add basic documentation for multi node training (#1988)

SumanthRH committed 2 years ago

Verified 936d2f4f

[docs] Quick tour refactor (#2008)

MKhalusova committed 2 years ago

Verified da98d601

fix resuming from checkpoint (#2001)

SumanthRH committed 2 years ago

Verified 658492fb

Older