huggingface/accelerate

Pull Requests Commits

Gate NUMA workaround on torch.numa presence, not torch>=2.10

SunMarc committed 20 hours ago

e7824bc0

fix notebook launcher cuda init

SunMarc committed 4 days ago

b86f2a31

fix: add missing neuron device case (#4042)

michaelbenayoun committed 16 days ago

Verified 1fab6edd

[CI] Bump style-bot SHA + switch to GitHub App (#4031)

paulinebm committed 26 days ago

Verified 29e03d18

Fix region compilation fsdpv2 (#4022)

SunMarc committed 34 days ago

Verified 917e2a9e

Fix fsdp2 load full state dict dtype mismatch (#4021)

SunMarc committed 36 days ago

Verified d1d2bba4

Fsdp2 fully_shard embedding and norm (#4015)

SunMarc committed 39 days ago

Verified 60b5c258

chore: bump doc-builder SHA for main doc build workflow (#4018)

rtrompier committed 39 days ago

Verified 95aaf22e

[FSDP2] Cast model to uniform dtype before fully_shard to fix mixed-dtype AssertionError (#3985)

roycho96 committed 40 days ago

Verified 92cc533e

Prevent double-wrapping models in prepare_model() (#3977)

joshuaswanson committed 46 days ago

Verified f1e6505b

chore: update doc-builder workflow SHA (#4009)

rtrompier committed 48 days ago

Verified 47d66d2a

[FSDP2] Auto-exclude non-floating frozen Params4bit from fully_shard to prevent QLoRA crash (#3987)

roycho96 committed 53 days ago

Verified 5adf855f

Add padded allgather and allreduce to reduce recompilation for Neuron devices. (#4000)

czkkkkkk committed 53 days ago

Verified 1bffb53d

fix: pass step parameter in TrackioTracker.log() (#3970)

liuyun7345 committed 54 days ago

Verified 94f975df

fix(FSDP2): auto-wrap policy ignoring _no_split_modules fallback (#3999)

JohnGiorgi committed 54 days ago

Verified b6f97773

Relax numerical tolerance for XPU in test_big_modeling (#4001)

YangKai0616 committed 54 days ago

Verified cff24503

pytorch-triton-xpu rename to triton-xpu (#4007)

sywangyi committed 54 days ago

Verified 072499ae

🔒 Pin GitHub Actions to commit SHAs (#3992)

paulinebm committed 55 days ago

Verified 01d097ae

fix: use key-based matching in fsdp2_load_full_state_dict (#3982)

roycho96 committed 71 days ago

Verified 0d8ca1a4

Fix TrackioTracker.log() ignoring step parameter (#3975)

joshuaswanson committed 74 days ago

Verified 1622df33

fix: add missing model_has_params4bit guard to fsdp2_load_full_state_dict call (#3981)

roycho96 committed 74 days ago

Verified 3936d308

Fix DistributedType documentation (#3980)

3manifold committed 74 days ago

Verified 71e8551f

Fix grammar, spelling, and consistency issues across docs and examples (#3961)

cihandemir committed 74 days ago

Verified 858ed553

Fix is_hf_initialized attribute (#3976)

SunMarc committed 77 days ago

Verified 13640d32

Torchao offload (#3973)

SunMarc committed 81 days ago

Verified 3475dcfa

fix: conditionally import torch.distributed.algorithms.join in accelerator (#3962)

0xDELUXA committed 83 days ago

Verified 9bbee85f

allow kernels flash-attn (#3959)

kashif committed 84 days ago

Verified f317f824

Fix iterable dataset sharding condition when n_shards == num_processes (#3958)

SunMarc committed 88 days ago

Verified 3e27a7c6

SunMarc committed 90 days ago

74427420

Fix testing ci (#3955)

SunMarc committed 90 days ago

Verified 2a7e27f7

Older