huggingface/accelerate

Pull Requests Commits

test extract parallel model

IlyasMoutawwakil committed 347 days ago

b65a26f0

default is True

IlyasMoutawwakil committed 353 days ago

23f95bb7

safe inplace regional compilation as a deafult

IlyasMoutawwakil committed 353 days ago

bcc3a2ec

revert repeated class check

IlyasMoutawwakil committed 353 days ago

f351bbd1

IlyasMoutawwakil committed 353 days ago

325e80ce

IlyasMoutawwakil committed 353 days ago

9d7efcaa

IlyasMoutawwakil committed 353 days ago

1eb4a956

IlyasMoutawwakil committed 353 days ago

c7388024

IlyasMoutawwakil committed 353 days ago

de3dfa15

remove unecessary pops

IlyasMoutawwakil committed 353 days ago

ac0e5658

make compile regions inplace to avoid recursion in some scenarios

IlyasMoutawwakil committed 353 days ago

386d874b

IlyasMoutawwakil committed 353 days ago

53f33be4

IlyasMoutawwakil committed 353 days ago

39bdbc80

Fix: Defer Tracker Initialization to Prevent Premature Distributed Setup (#3581)

yuanjua committed 355 days ago

Verified db536cbf

enable regional_compilation benchmark on xpu (#3592)

yao-matrix committed 355 days ago

Verified 4e9d0deb

Add kwargs to optimizer, scheduler and dataloader using function `accelerator().load_state()` (#3540)

luiz0992 committed 359 days ago

Verified 8cb3ace8

Resolve logger warnings (#3582)

emmanuel-ferdman committed 359 days ago

Verified b6d97cb8

Add support for standalone mode when default port is occupied on single node (#3576)

laitifranz committed 361 days ago

Verified 33967d47

enable test_cli & test_example cases on XPU (#3578)

yao-matrix committed 361 days ago

Verified 5b1fcda3

goodbye torch_ccl (#3580)

yao-matrix committed 361 days ago

Verified f55f0533

enable test_load_checkpoint_and_dispatch_with_broadcast cases on XPU (#3579)

yao-matrix committed 362 days ago

Verified 1ec99f0b

bump to v1.8.0dev

SunMarc committed 1 year ago

417bc529

enable test_dispatch_model_tied_weights_memory_with_nested_offload_cpu on xpu (#3569)

yao-matrix committed 1 year ago

Verified 97c93c48

set backend correctly for CUDA+FSDP2+cpu-offload (#3574)

SunMarc committed 1 year ago

Verified cd37bbb6

Fix prevent duplicate GPU usage in distributed processing (#3526)

ved1beta committed 1 year ago

Verified 7aa3b56c

reenable FSDP2+qlora support (#3546)

winglian committed 1 year ago

Verified 14f4306c

Add regional compilation to cli tools and env vars (#3572)

IlyasMoutawwakil committed 1 year ago

Verified e6e71758

tune env command output (#3570)

yao-matrix committed 1 year ago

Verified 1f6efcea

simplify model.to logic (#3562)

yao-matrix committed 1 year ago

Verified 9fa97f96

add xpu synchronize (#3563)

faaany committed 1 year ago

Verified 764eee4a

Older