microsoft/DeepSpeed

Pull Requests Commits

Remove unnecessary device synchronization for stage 2 (#2500)

li-yi-dong committed 2 years ago

Verified 97deaaec

Abstract accelerator (step 2) (#2560)

delock committed 2 years ago

Verified 9548d48f

Fix Opt injection (#2541)

RezaYazdaniAminabadi committed 2 years ago

Verified 95d9a1b6

[launcher] fail gracefully if hostname -i doesn't work as expected (#2631)

jeffra committed 2 years ago

Verified a091bc22

[doc] fix `min_loss_scale` default (#2660)

stas00 committed 2 years ago

Verified f2ea7a38

tweaks to ds-attn, distilbert policy, and mup (#2649)

jeffra committed 2 years ago

Verified d9b788d7

fix assertion error in zero stage 3 (#2647)

GuanhuaWang committed 2 years ago

Verified 6375cb3f

Fix issue w. bloom when changing tp size (#2645)

jeffra committed 2 years ago

Verified e0aa84c5

[inference] ds-attention refactor w.r.t. ops (#2623)

jeffra committed 2 years ago

Verified bb68c526

[zero-3] Handle forward parameter return correctly in nested cases (#2642)

samyam committed 2 years ago

Verified a298a43a

Fix typo in autotuner.py (#2639)

eltociear committed 2 years ago

Verified 6273dffc

add enable_each_rank_log to deepspeed/launcher/runner.py (#2571)

inkcherry committed 2 years ago

Verified 11f5daba

Unit tests setup own venv (#2628)

mrwyattii committed 2 years ago

Verified 6fff50f5

[launcher] parse hostfile via regex and added error checks (#2626)

jeffra committed 2 years ago

Verified 8c56c25d

[inference] check for unsupported model generate args (#2627)

jeffra committed 2 years ago

Verified 5676f5ec

Add Megatron CI workflow (#2614)

mrwyattii committed 2 years ago

Verified df985fac

Update AVX512 Detection (#2621)

cmikeh2 committed 2 years ago

Verified a25c31b6

fixes #2498 (#2603)

Alexander Jipa committed 2 years ago

Verified 0f0e38c5

Remove GatheredParameters context from replace_with_policy (#2591)

lekurile committed 2 years ago

Verified 503706ac

call empty_cache to really free up GPU memory as described in comment (#2620)

guoyejun committed 2 years ago

Verified d0dbc95a

skip torch.zeros and tensor.copy_ when model parallel is not used (#2479)

guoyejun committed 2 years ago

Verified b08cf416

Use rocm/pytorch:latest (#2613)

jithunnair-amd committed 2 years ago

Verified 7e2103f8

add fix for older pydantic versions (#2611)

mrwyattii committed 2 years ago

Verified 2e596e67

[deepspeed/autotuner] Bug fix for binary search for batch size (#2162)

rahilbathwal5 committed 2 years ago

Verified 384f17b0

Move layer norm to new schedule (#2590)

lokoppakmsft committed 2 years ago

Verified 3a3dfe66

Migrate ops tests to new inference_ops marker (#2599)

cmikeh2 committed 2 years ago

Verified 7425a365

fix blog link (#2600)

conglongli committed 2 years ago

Verified acde873c

DeepSpeed Data Efficiency Library (#2585)

conglongli committed 2 years ago

Verified ef869377

jeffra committed 2 years ago

2600db54

get mask token from tokenizer (#2592)

jeffra committed 3 years ago

Verified 2076bf23

Newer Older