microsoft/DeepSpeed

Pull Requests Commits

Merge branch 'tohtana/cache_kv_requirements' of github.com:microsoft/DeepSpeed into tohtana/cache_kv_requirements

Masahiro Tanaka committed 1 year ago

11ddfcde

pass cache size to lru_cache

Masahiro Tanaka committed 1 year ago

73738157

Merge branch 'master' into tohtana/cache_kv_requirements

mrwyattii committed 1 year ago

Verified 45e4a3a6

Optimize preprocess for ragged batching (#4942)

tohtana committed 1 year ago

Verified d0b238a3

Merge branch 'tohtana/cache_kv_requirements' of github.com:microsoft/DeepSpeed into tohtana/cache_kv_requirements

Masahiro Tanaka committed 1 year ago

ed9194df

disable cache for old python

Masahiro Tanaka committed 1 year ago

dadab62e

Merge branch 'master' into tohtana/cache_kv_requirements

tohtana committed 1 year ago

Verified ac31cd31

cache kv requirements

Masahiro Tanaka committed 1 year ago

f42d8215

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

merge branch 'tohtana/optimize_form_batch' of github.com:microsoft/DeepSpeed into tohtana/optimize_form_batch

Masahiro Tanaka committed 1 year ago

33431948

add comments to clarify the intention

Masahiro Tanaka committed 1 year ago

3c0ab3d7

Update test_inference.py

mrwyattii committed 1 year ago

Verified a7260066

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

loadams committed 1 year ago

Verified 05cc3462

optimize preprocess for ragged batching

Masahiro Tanaka committed 1 year ago

a4d0fd78

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

Older