huggingface/accelerate

Pull Requests Commits

Release: v0.30.1

muellerzr committed 1 year ago

66eefd7f

fix fstr format (#2810)

Jintao-Huang committed 1 year ago

Verified 83bad875

Optimize the megatron plugin (#2822)

zhangsheng377 committed 1 year ago

Verified 24d8b63f

monitor-interval, take 2 (#2833)

muellerzr committed 1 year ago

Verified 4a83ee53

Improve test speeds by up to 30% in multi-gpu settings (#2830)

muellerzr committed 1 year ago

Verified 05d240af

Fix DeepSpeed config validation error by changing `stage3_prefetch_bucket_size` value to an integer (#2814)

adk9 committed 1 year ago

Verified bad2ce42

Remove out-dated xpu device check code in `get_balanced_memory` (#2826)

faaany committed 1 year ago

Verified 30cb7ece

add cuda dep for a test (#2820)

SunMarc committed 1 year ago

Verified b7fa2fa9

State dictionary retrieval from offloaded modules (#2619)

blbadger committed 1 year ago

Verified d5d378d6

4-bit quantization meta device bias loading bug (#2805)

SunMarc committed 1 year ago

Verified 065e74d1

Fix access error for torch.mps when using torch==1.13.1 on macOS (#2806)

SunMarc committed 1 year ago

Verified 86b6deae

New template (#2808)

muellerzr committed 1 year ago

Verified b24a0ef5

fix comet test (#2804)

SunMarc committed 1 year ago

Verified e061edc6

Fix type in accelerator.py (#2800)

qgallouedec committed 1 year ago

Verified c3f42269

Fix Wrong use of sync_gradients used to implement sync_each_batch (#2790)

fabianlim committed 1 year ago

Verified 05534836

Add Elastic Launch Support to `notebook_launcher` (#2788)

yhna940 committed 1 year ago

Verified 415789d0

fix duplicate elements in split_between_processes (#2781)

hkunzhe committed 1 year ago

Verified ae472bac

Fixup CLI test (#2796)

muellerzr committed 1 year ago

Verified 4f2c2ba4

Upgrade huggingface's megatron to nvidia's megatron when use MegatronLMPlugin (#2501)

zhangsheng377 committed 1 year ago

Verified e26065a2

FIX / FSDP : Guard fsdp utils for earlier PyTorch versions (#2794)

younesbelkada committed 1 year ago

Verified 1cb6fdcf

Introduce shard-merging util for FSDP (#2772)

muellerzr committed 1 year ago

Verified 4ba436ec

Skip tied weights disk offload test (#2782)

SunMarc committed 1 year ago

Verified 91e8a3ce

Add arg from CLI to fix failing test (#2783)

muellerzr committed 1 year ago

Verified 4ad4d28c

Enable config for fsdp activation checkpointing (#2779)

helloworld1 committed 1 year ago

Verified befd87f0

Skip deepspeed test (#2776)

SunMarc committed 1 year ago

Verified abce3604

Fix small edge case in get_module_leaves (#2774)

SunMarc committed 1 year ago

Verified 27a607ea

fix minor typo (#2767)

TemryL committed 1 year ago

Verified aa21174d

optimize get_module_leaves speed (#2756)

BBuf committed 1 year ago

Verified 6cf1cc0a

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

muellerzr committed 1 year ago

Verified bb465a9c

Enable sharded cpu resume (#2762)

muellerzr committed 1 year ago

Verified 67308ca6

Older