huggingface/accelerate

Pull Requests Commits

Update big_modeling.md for layerwise casting (#3548)

sayakpaul committed 332 days ago

Verified d02e51cc

Fix CI due to missing package (#3535)

SunMarc committed 339 days ago

Verified c5caa11e

Update Docker builds to align with CI requirements (#3532)

matthewdouglas committed 340 days ago

Verified 39e2bebb

Fix logic in `accelerator.prepare` + IPEX for 2+ `nn.Models` and/or `optim.Optimizers` (#3517)

mariusarvinte committed 343 days ago

Verified 0af45bf1

[FSDP2] Issues in Wrap Policy and Mixed Precision (#3528)

jhliu17 committed 343 days ago

Verified 806ac848

[FSDP2] Fix memory spike with `cpu_ram_efficient_loading=True` (#3482)

S1ro1 committed 344 days ago

Verified 23b09250

[FSDP2] Enable FULL_STATE_DICT (#3527)

S1ro1 committed 345 days ago

Verified 8fb07353

Solve link error in internal_mechanism documentation (#3506) (#3507)

alvaro-mazcu committed 345 days ago

Verified 4f35cf71

fix cuda init (#3530)

SunMarc committed 345 days ago

Verified ada21cfb

Add torchao to FP8 error message (#3514)

jphme committed 346 days ago

Verified b451956f

[Feat] Layerwise casting hook (#3427)

sayakpaul committed 346 days ago

Verified 6a9a6152

fix cache (#3513)

SunMarc committed 350 days ago

Verified 423fbbfd

Remove deprecated PyTorch/XLA APIs (#3484)

zpcore committed 353 days ago

Verified 34c17798

Fix: require transformers version for tp tests (#3504)

S1ro1 committed 353 days ago

Verified 54496571

fix: apply torchfix to set `weights_only=True` (#3497)

bzhong-solink committed 353 days ago

Verified 4a3cbcb6

Add FP8 runners + tweak building FP8 image (#3493)

zach-huggingface committed 353 days ago

Verified 583b26db

Fix deepspeed tests (#3503)

S1ro1 committed 354 days ago

Verified 7812d979

(Part 1) fix: make TP training compatible with new transformers (#3457)

kmehant committed 357 days ago

Verified 67adb473

nit: needed sanity checks for fsdp2 (#3499)

kmehant committed 357 days ago

Verified ee4cab96

Use `torch.distributed.checkpoint.state_dict.set_model_state_dict` in `load_checkpoint_in_model` (#3432)

Matthew Hoffman committed 357 days ago

Verified 73c2378c

Add the HPU into accelerate config (#3495)

yuanwu2017 committed 358 days ago

Verified b2f937fa

[bug] unsafe_serialization option doesn't work (#3496)

cyr0930 committed 359 days ago

Verified 3b899877

fix warning error (#3491)

faaany committed 359 days ago

Verified a43e4170

fix fp8 config (#3492)

SunMarc committed 359 days ago

Verified 334d6ab9

add support for custom function for reducing the batch size (#3071)

winglian committed 360 days ago

Verified 650b6659

Don't create new param for TorchAO sequential offloading due to weak BC guarantees (#3444)

a-r-r-o-w committed 360 days ago

Verified fb909963

Fix check_tied_parameters_in_config for multimodal models (#3479)

SunMarc committed 360 days ago

Verified 32b2e160

Update low_precision_training.md (#3488)

sadra-barikbin committed 360 days ago

Verified 8c0a2962

Adds style bot (#3478)

zach-huggingface committed 1 year ago

Verified 63168b15

use device agnostic torch.OutOfMemoryError from pytorch 2.5.0 (#3475)

yao-matrix committed 1 year ago

Verified 3cf5e4c8

Newer Older