huggingface/accelerate

Pull Requests Commits

Improve deepspeed env gen (#2565)

muellerzr committed 2 years ago

Verified c2f193a2

Add log message for RTX 4000 series when performing multi-gpu inference with device_map (#2557)

SunMarc committed 2 years ago

Verified 18121523

Add NUMA affinity control for NVIDIA GPUs (#2535)

muellerzr committed 2 years ago

Verified b8b353b7

Add Cambricon MLU accelerator support (#2552)

huismiling committed 2 years ago

Verified f2778d65

🚨🚨🚨Move to using tags rather than latest for docker images and consolidate image repos 🚨 🚨🚨 (#2554)

muellerzr committed 2 years ago

Verified 2ad42e77

Include working driver check (#2558)

muellerzr committed 2 years ago

Verified e8aaee5d

split_between_processes for Dataset (#2433)

geronimi73 committed 2 years ago

Verified 910c1b6a

Add mapping main_process_ip and ip-master_addr when not using standard as deepspeed launcher (#2495)

asdfry committed 2 years ago

Verified 92d3240b

Fix test_script.py on TPU v2/v3 (#2542)

vanbasten23 committed 2 years ago

Verified 02a8a9a3

muellerzr committed 2 years ago

ee163b66

Use uv instead of pip install for github CI (#2546)

muellerzr committed 2 years ago

Verified 354db5b5

Update FSDP mixed precision setter to enable fsdp+qlora (#2544)

pacman100 committed 2 years ago

Verified 92b1ad01

Allow Gradients to be Synced Each Data Batch While Performing Gradient Accumulation (#2531)

fabianlim committed 2 years ago

Verified 60bfdaa9

Remove extra double-dash in error message (#2541)

drscotthawley committed 2 years ago

Verified 16eb6d76

[docs] Troubleshoot (#2538)

stevhliu committed 2 years ago

Verified c8acfa70

Overdue email change... (#2534)

muellerzr committed 2 years ago

Verified e70e3c87

vmoens committed 2 years ago

Verified bc8dfe3c

Check if the buffers fit GPU memory after device map auto inferred (#2412)

notsyncing committed 2 years ago

Verified e3d32424

Update link to dynamo/compile doc (#2533)

thethomasboyer committed 2 years ago

Verified 10882eed

Update the default behavior of `zero_grad(set_to_none=None)` (#2472)

yongchanghao committed 2 years ago

Verified 145a98fc

Enable using dash or underscore for CLI args (#2527)

muellerzr committed 2 years ago

Verified 64ae9ea3

Launch mpirun from accelerate launch for multi-CPU training (#2493)

dmsuehir committed 2 years ago

Verified 8aa72b97

Remove unnecessary `env=os.environ.copy()`s (#2449)

akx committed 2 years ago

Verified 97d115a2

qbitstensor compatibility (#2526)

SunMarc committed 2 years ago

Verified 63cfd9ef

Don't manage `PYTORCH_NVML_BASED_CUDA_CHECK` when calling `accelerate.utils.imports.is_cuda_available()` (#2524)

luiscape committed 2 years ago

Verified 6cf8221a

Add copyright + some ruff lint things (#2523)

muellerzr committed 2 years ago

Verified 7a2feeca

fix typo in launch.py (#2516)

DerrickWang005 committed 2 years ago

Verified ee004674

[docs] Fix typos (#2490)

omahs committed 2 years ago

Verified 65544d8f

Fix edge case in infer_auto_device_map when dealing with buffers (#2511)

SunMarc committed 2 years ago

Verified 5fce525f

Fixed 0MiB bug in convert_file_size_to_int (#2507)

StoyanStAtanasov committed 2 years ago

Verified ca37b0e4

Newer Older