Pull Requests huggingface/accelerate

enable big_model_inference on xpu

#3595 opened 2025-05-27 07:29 by yao-matrix

Update Gaudi Runners

#3593 opened 2025-05-26 13:46 by IlyasMoutawwakil

Remove device_count for TPU launcher to avoid initializing runtime

#3587 opened 2025-05-24 16:18 by sorgfresser

[FSDP2, Do not merge] FP8 adjustments

#3585 opened 2025-05-22 14:27 by S1ro1

feat: use datasets.IterableDataset shard if possible.

#3583 opened 2025-05-22 09:11 by ValMystletainn

Add CustomTypesDataLoader for handling any iterable dataloader

#3519 opened 2025-04-19 05:26 by ved1beta

WIP: Compose TP + DDP/FSDP2

#3498 opened 2025-04-10 16:10 by S1ro1

(fix) remove sampler_is_batch_sampler code in prepare_data_loader(..)

#3469 opened 2025-03-31 13:10 by suzyahyah

Add bf16/fp16 support for amp with mps device wip

#3373 opened 2025-01-29 14:34 by SunMarc

[WIP] optimize infer_auto_device_map for multi-GPU allocation wip

#3321 opened 2025-01-02 20:05 by Nech-C

Some adjustment for supporting Deepspeed-Ulysses wip

#2877 opened 2024-06-20 16:50 by zeyugao

handle weight sharing with init_on_device wip

#2737 opened 2024-05-03 01:07 by aws-rhsoln

[DO NOT MERGE] add all level buffer support when computing infer_auto_device_map enhancement feature request wip

#2663 opened 2024-04-12 12:52 by SunMarc

Enable cpu offload with weights inside the module wip

#2214 opened 2023-12-04 23:11 by SunMarc