huggingface/accelerate

Pull Requests Commits

Feat: better example script

S1ro1 committed 243 days ago

2b8e81f5

Initial api design

S1ro1 committed 246 days ago

6cdc6d87

S1ro1 committed 247 days ago

2a13375c

S1ro1 committed 247 days ago

400fe57d

SunMarc committed 247 days ago

Verified 31af8d4e

Add support for e5e2 and default to hybrid when launcher is used (#3640)

IlyasMoutawwakil committed 247 days ago

Verified b7493a82

bump to v1.9.0dev

SunMarc committed 248 days ago

a16d2bb3

fix grad acc deepspeed (#3638)

SunMarc committed 248 days ago

Verified cac22ed9

Fix: correct labels (#3637)

S1ro1 committed 248 days ago

Verified be826a6b

Feat: add cpu offload (#3636)

S1ro1 committed 249 days ago

Verified 59396408

[DeepSpeed] sync gradient accum steps from deepspeed plugin (#3632)

kashif committed 249 days ago

Verified 7f9c8cbe

feat: use datasets.IterableDataset shard if possible (#3635)

SunMarc committed 249 days ago

Verified 9888c7ed

Fix Typos in Documentation and Comments (#3621)

leopardracer committed 249 days ago

Verified 42a68c30

Integrate SwanLab for offline/online experiment tracking for Accelerate (#3605)

ShaohonChen committed 249 days ago

Verified 6597dae7

remove hardcoded cuda from fsdpv2 (#3631)

IlyasMoutawwakil committed 250 days ago

Verified 8878d937

remove ipex.optimize in accelerate (#3608)

yao-matrix committed 250 days ago

Verified 2eaf5cdb

[Deepspeed] deepspeed auto grad accum (#3630)

kashif committed 251 days ago

Verified 23c1d8db

add xpu support in TorchTensorParallelPlugin (#3627)

yao-matrix committed 254 days ago

Verified 0af621bb

Add fp8_e5m2 support in `dtype_byte_size` (#3625)

SunMarc committed 255 days ago

Verified bee04f1b

fix xpu 8bit value loading (#3623)

jiqing-feng committed 255 days ago

Verified 8a953f08

small fix (#3619)

IlyasMoutawwakil committed 256 days ago

Verified 3518c035

Remove device_count (#3587)

sorgfresser committed 257 days ago

Verified 2f8fd72e

[FSDP2] Refactor + FP8 (#3585)

S1ro1 committed 257 days ago

Verified d2e6b031

better handle FP8 with and without deepspeed (#3611)

IlyasMoutawwakil committed 257 days ago

Verified b9fee48c

Fix bf16 training with TP (#3610)

SunMarc committed 257 days ago

Verified 3a82b056

fix deepspeed regional compilation (#3609)

IlyasMoutawwakil committed 261 days ago

Verified 6b61a373

Update Gaudi Runners (#3593)

IlyasMoutawwakil committed 264 days ago

Verified 682691de

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

S1ro1 committed 264 days ago

Verified 791055b4

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

yao-matrix committed 268 days ago

Verified 16bf1d89

enable big_model_inference on xpu (#3595)

yao-matrix committed 268 days ago

Verified ab3c604e

Older