microsoft/DeepSpeed

Pull Requests Commits

Fix docs inconsistency on default value for

loadams committed 1 year ago

0144c0bc

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

Nvme offload checkpoint (#4707)

eisene committed 1 year ago

Verified d058d4b3

fix num_kv_heads sharding in uneven autoTP for Falcon-40b (#4712)

Yejing-Lai committed 1 year ago

Verified 1787673e

[NPU]Add ZeRO-Infinity feature for NPU (#4809)

misstek committed 1 year ago

Verified b596963b

Release overlap_comm & contiguous_gradients restrictions for ZeRO 1 (#4887)

li-plus committed 1 year ago

Verified af033831

enable starcode((kv_head=1)) autotp (#4896)

Yejing-Lai committed 1 year ago

Verified 85132adc

[XPU] XPU accelerator support for Intel GPU device (#4547)

delock committed 1 year ago

Verified f4f31317

add sharded loading for safetensors in AutoTP (#4854)

sywangyi committed 1 year ago

Verified c8c57b8c

Support cpu tensors without direct device invocation (#3842)

abhilash1910 committed 1 year ago

Verified c84c28d2

Add support of Microsoft Phi-2 model to DeepSpeed-FastGen (#4812)

arashb committed 1 year ago

Verified 83427253

Older