deepspeedai/DeepSpeed

Pull Requests Commits

Update phi-2 InferenceV2 model type

lekurile committed 1 year ago

10f322af

[MiCS]Add the path to support sequence_data_parallel on MiCS (#4926)

ys950902 committed 1 year ago

Verified 752a50ac

Stage_1_and_2.py: fix assert for reduce_scatter configurations combinations (#4964)

nelyahu committed 1 year ago

Verified 69a45988

Add missing op_builder.hpu component for HPU accelerator (#4963)

nelyahu committed 1 year ago

Verified e2780764

[docs] Add new autotp supported model in tutorial (#4960)

delock committed 1 year ago

Verified 7739c0ac

Fix bug where ZeRO2 never uses the reduce method. (#4946)

CurryRice233 committed 1 year ago

Verified a85b6e47

Optimize preprocess for ragged batching (#4942)

tohtana committed 1 year ago

Verified d0b238a3

fix uneven issue & add balance autotp (#4697)

Yejing-Lai committed 1 year ago

Verified 29417ab5

Fix bug in CI model caching (#4951)

mrwyattii committed 1 year ago

Verified 43daf413

Fix docs inconsistency on default value for `ignore_unused_parameters` (#4949)

loadams committed 1 year ago

Verified 05cc3462

Cache HF model list for inference tests (#4940)

mrwyattii committed 1 year ago

Verified 13d84b49

InfV2 - remove generation config requirement (#4938)

mrwyattii committed 1 year ago

Verified 62004e0d

Only run MII CI for inference changes (#4939)

mrwyattii committed 1 year ago

Verified c6b975b8

Allow specifying mii branch for nv-a6000 workflow (#4936)

mrwyattii committed 1 year ago

Verified 47389852

autoTP for Qwen (#4902)

inkcherry committed 1 year ago

Verified ee7db483

Enable quantizer op on ROCm (#4114)

rraminen committed 1 year ago

Verified f2d4845f

Specify permissions for secrets.GITHUB_TOKEN (#4927)

mrwyattii committed 1 year ago

Verified 6112d08c

Fix confusing width in simd_load (#4714)

yzhblind committed 1 year ago

Verified f5179031

DeepSpeedZeroOptimizer: refactor bit16 flattening to support more accelerators (#4833)

nelyahu committed 1 year ago

Verified ade98365

Add support of Qwen models (7b, 14b, 72b) to DeepSpeed-FastGen (#4913)

ZonePG committed 1 year ago

Verified ed10cc73

Pin to triton 2.1.0 to fix issues with nv-inference (#4929)

loadams committed 1 year ago

Verified acdf570a

Refactor the positional emebdding config code (#4920)

arashb committed 1 year ago

Verified c1e02052

fix falcon-40b accuracy issue (#4895)

Yejing-Lai committed 1 year ago

Verified 16c265c0

Change PR/schedule triggers for CPU-inference (#4924)

loadams committed 1 year ago

Verified 43eba775

Add paths to run SD tests (#4919)

loadams committed 1 year ago

Verified 5457363c

[Fix] Fix cpu inference UT failure (#4430)

delock committed 1 year ago

Verified d8d865f4

Fix SD workflow to work with latest diffusers version (#4918)

lekurile committed 1 year ago

Verified 75db3d7d

Unit tests for MiCS (#4792)

zarzen committed 1 year ago

Verified d7b764e3

Add Habana Labs HPU accelerator support (#4912)

deepcharm committed 1 year ago

Verified 838c69be

Add WarmupCosineLR to Read the Docs (#4916)

dwyatte committed 1 year ago

Verified 59c5f37e

Older