microsoft/DeepSpeed

Pull Requests Commits

Add Compressedbackend for Onebit optimizers (#5473)

Liangliang-Ma committed 1 year ago

Verified 11a62a06

Merge branch 'master' into costineseanu/windows_inference_build

costin-eseanu committed 1 year ago

Verified edf5780e

Merge branch 'master' into costineseanu/windows_inference_build

costin-eseanu committed 1 year ago

Verified 6f655d49

Merge branch 'costineseanu/windows_inference_build' of https://github.com/microsoft/DeepSpeed into costineseanu/windows_inference_build

Costin Eseanu committed 1 year ago

72828354

Fixed Windows setup.py not copying files in the right location.

Costin Eseanu committed 1 year ago

a6120c85

fix sequence parallel(Ulysses) grad scale for zero0 (#5555)

inkcherry committed 1 year ago

Verified 6b6d6418

Fix cuda hardcode for inference woq (#5565)

Liangliang-Ma committed 1 year ago

Verified af4356b3

Merge branch 'master' into costineseanu/windows_inference_build

costin-eseanu committed 1 year ago

Verified 58a62a8c

estimate_zero2_model_states_mem_needs: fixing memory estiamtion (#5099)

nelyahu committed 1 year ago

Verified f4cb866c

Fixed whitespace.

Costin Eseanu committed 1 year ago

c2496f9d

Fixed the Windows build for inference ops.

Costin Eseanu committed 1 year ago

25150c1e

Fixed the Windows build. (#5596)

costin-eseanu committed 1 year ago

Verified e7dd28a2

Add slide deck for meetup in Japan (#5598)

tohtana committed 1 year ago

Verified 77c94942

Update minor CUDA version compatibility (#5591)

adk9 committed 1 year ago

Verified 1baf6884

DeepSpeedCheckpoint: support custom final ln idx (#5506)

nelyahu committed 1 year ago

Verified 2fc702ed

Update to fix sidebar over text (#5567)

loadams committed 1 year ago

Verified 4deb40de

Rename files in fp_quantize op from quantize.* to fp_quantize.* (#5577)

loadams committed 1 year ago

Verified bf66acdb

[MiCS] Remove the handle print on DeepSpeed side (#5574)

ys950902 committed 1 year ago

Verified fd8051a6

Update HPU docker version (#5566)

loadams committed 1 year ago

Verified 988372b7

Add hybrid_engine.py as path to trigger the DS-Chat GH workflow (#5562)

lekurile committed 1 year ago

Verified 3a3a6db3

Add Ulysses DistributedAttention compatibility (#5525)

Kwen-Chen committed 1 year ago

Verified f86824be

Add throughput timer configuration (#5363)

deepcharm committed 1 year ago

Verified 995ba119

[INF] DSAttention allow input_mask to have false as value (#5546)

oelayan7 committed 1 year ago

Verified f4efef21

Update to HF_HOME from TRANSFORMERS_CACHE (#4816)

loadams committed 1 year ago

Verified 263bfe28

Adapt doc for #4405 (#5552)

oraluben committed 1 year ago

Verified 29903925

Small typos in functions set_none_gradients_to_zero (#5557)

TravelLeraLone committed 1 year ago

Verified 975290ae

Avoid overwrite of compiled module wrapper attributes (#5549)

deepcharm committed 1 year ago

Verified 5b314f4e

Remove synchronize calls from allgather params (#5516)

Liran Bachar committed 1 year ago

Verified 0a174038

Fix RuntimeError for moe on XPU: tensors found at least two devices (#5519)

LevelDownRefine committed 1 year ago

Verified 695d79ea

Fix the TypeError for XPU Accelerator (#5531)

LevelDownRefine committed 1 year ago

Verified 1d819673

Newer Older