microsoft/DeepSpeed

Pull Requests Commits

Fix hpZ with zero element (#5652)

samadejacobs committed 1 year ago

Verified b33873d2

Remove compile wrapper to simplify access to model attributes (#5581)

tohtana committed 1 year ago

Verified 2a0c0e3c

fix IDEX dependence in xpu accelerator (#5666)

Liangliang-Ma committed 1 year ago

Verified 8831b57b

[XPU] adapt lazy_call func to different versions (#5670)

YizhouZ committed 1 year ago

Verified 73316307

_exec_forward_pass: place zeros(1) on the same device as the param (#5576)

nelyahu committed 1 year ago

Verified 05cb79db

[CPU] SHM based allreduce improvement for small message size (#5571)

delock committed 1 year ago

Verified eda5075b

Update version.txt after 0.14.3 release (#5651)

mrwyattii committed 1 year ago

Verified dfcade24

Monitor was always enabled causing performance degradation (#5633)

deepcharm committed 1 year ago

Verified 54f98fd9

pipe/_exec_backward_pass: fix immediate grad update (#5605)

nelyahu committed 1 year ago

Verified 91a9986d

assumption of torch.initial_seed function accepting seed arg in DeepSpeedAccelerator abstract class is incorrect (#5569)

polisettyvarma committed 1 year ago

Verified ac935c7f

fixes in _partition_param_sec function (#5613)

mmhab committed 1 year ago

Verified b6e24adb

Fix overlap communication of ZeRO stage 1 and 2 (#5606)

penn513 committed 1 year ago

Verified a41729f6

stage_1_and_2: optimize clip calculation to use clamp (#5632)

nelyahu committed 1 year ago

Verified 1ef9b029

WA for Torch-compile-Z3-act-apt accuracy issue from the Pytorch repo (#5590)

NirSonnenschein committed 1 year ago

Verified 6e2899fb

Pin transformers version for MII tests (#5629)

loadams committed 1 year ago

Verified 31a57fa3

Updated hpu-gaudi2 tests content. (#5622)

vshekhawat-hlab committed 1 year ago

Verified 69adeab6

Add Compressedbackend for Onebit optimizers (#5473)

Liangliang-Ma committed 1 year ago

Verified 11a62a06

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

estimate_zero2_model_states_mem_needs: fixing memory estiamtion (#5099)

nelyahu committed 1 year ago

Verified f4cb866c

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

Newer Older