huggingface/accelerate

Pull Requests Commits

add xpu support in TorchTensorParallelPlugin (#3627)

yao-matrix committed 248 days ago

Verified 0af621bb

Add fp8_e5m2 support in `dtype_byte_size` (#3625)

SunMarc committed 249 days ago

Verified bee04f1b

fix xpu 8bit value loading (#3623)

jiqing-feng committed 249 days ago

Verified 8a953f08

small fix (#3619)

IlyasMoutawwakil committed 250 days ago

Verified 3518c035

Remove device_count (#3587)

sorgfresser committed 251 days ago

Verified 2f8fd72e

[FSDP2] Refactor + FP8 (#3585)

S1ro1 committed 251 days ago

Verified d2e6b031

better handle FP8 with and without deepspeed (#3611)

IlyasMoutawwakil committed 251 days ago

Verified b9fee48c

Fix bf16 training with TP (#3610)

SunMarc committed 251 days ago

Verified 3a82b056

fix deepspeed regional compilation (#3609)

IlyasMoutawwakil committed 255 days ago

Verified 6b61a373

Update Gaudi Runners (#3593)

IlyasMoutawwakil committed 258 days ago

Verified 682691de

Fix: list object has no attribute keys (#3603)

S1ro1 committed 258 days ago

Verified 791055b4

enable torchao and pippy test cases on XPU (#3599)

yao-matrix committed 262 days ago

Verified 16bf1d89

enable big_model_inference on xpu (#3595)

yao-matrix committed 262 days ago

Verified ab3c604e

enable fsdp2 benchmark on XPU (#3590)

yao-matrix committed 265 days ago

Verified 273799c8

add device-agnostic GradScaler (#3588)

yao-matrix committed 265 days ago

Verified 43526c5c

change to use torch.device (#3594)

yao-matrix committed 265 days ago

Verified 07f2392f

[docs] no hard-coded cuda in the ddp documentation (#3589)

faaany committed 265 days ago

Verified ee2f48c2

Set ccl and KMP param in simple launch (#3575)

jiqing-feng committed 266 days ago

Verified 4f3abb73

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

yuanjua committed 266 days ago

Verified db536cbf

enable regional_compilation benchmark on xpu (#3592)

yao-matrix committed 266 days ago

Verified 4e9d0deb

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

luiz0992 committed 270 days ago

Verified 8cb3ace8

Resolve logger warnings (#3582)

emmanuel-ferdman committed 270 days ago

Verified b6d97cb8

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

laitifranz committed 272 days ago

Verified 33967d47

enable test_cli & test_example cases on XPU (#3578)

yao-matrix committed 272 days ago

Verified 5b1fcda3

goodbye torch_ccl (#3580)

yao-matrix committed 272 days ago

Verified f55f0533

enable test_load_checkpoint_and_dispatch_with_broadcast cases on XPU (#3579)

yao-matrix committed 273 days ago

Verified 1ec99f0b

bump to v1.8.0dev

SunMarc committed 277 days ago

417bc529

enable test_dispatch_model_tied_weights_memory_with_nested_offload_cpu on xpu (#3569)

yao-matrix committed 277 days ago

Verified 97c93c48

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

SunMarc committed 277 days ago

Verified cd37bbb6

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

ved1beta committed 277 days ago

Verified 7aa3b56c

Newer Older