microsoft/DeepSpeed

Pull Requests Commits

resolve formatting issue

Reza Yazdani committed 5 years ago

17c53ecb

add module_inject directory

Reza Yazdani committed 5 years ago

e436ee99

moving the inject part to module_inject directory

Reza Yazdani committed 5 years ago

a0816018

fixing the parameter data-copy

Reza Yazdani committed 5 years ago

773aa214

ds_transformer injection code

Reza Yazdani committed 5 years ago

f2e74517

[build] build against installed cuda-11.1 while torch built w/ cuda-11.0 (#570)

stas00 committed 5 years ago

Verified ff58fa7e

Add compute capability 8.0 if on cuda 11+ (#572)

jeffra committed 5 years ago

Verified be33bea4

[engine] train should be able to get `mode` arg (#571)

stas00 committed 5 years ago

Verified 2d1f7c01

Add 'latest' checkpoint save/load support (#569)

jeffra committed 5 years ago

Verified 845921b3

[cifar tutorial] improve readability (#567)

stas00 committed 5 years ago

Verified 7a75f8b3

tracking optimizer step in cpu-adam when loading checkpoint (#564)

RezaYazdaniAminabadi committed 5 years ago

Verified 9f52a36f

supporting different hidden dimensions (#559)

RezaYazdaniAminabadi committed 5 years ago

Verified c78c29f9

[doc] typo fix and clarification (#563)

stas00 committed 5 years ago

Verified 17f36f1b

jeffra committed 5 years ago

c51fa65d

update manifest

jeffra committed 5 years ago

e4e20662

bump to 0.3.6 and fix manifest to include reqs (#561)

jeffra committed 5 years ago

Verified 73c3262d

Adds long_description to setup.py (#560)

Shaden Smith committed 5 years ago

Verified 60097136

jeffra committed 5 years ago

16313a96

Turn back on PP tests (#558)

jeffra committed 5 years ago

Verified eec44af1

Simplify dist init and only init if needed. (#553)

awan-10 committed 5 years ago

Verified 0e831e23

Deprecate client ability to disable gradient reduction (#552)

tjruwase committed 5 years ago

Verified 6e65c2cc

Update badges and CI name (#557)

jeffra committed 5 years ago

Verified 1ef5cd23

Switch to CI to GitHub Actions (#556)

jeffra committed 5 years ago

Verified 3347460e

Create main.yml

jeffra committed 5 years ago

Verified c18fb0de

Bug fix for norm calculation in absence of model parallel group (#551)

samyam committed 5 years ago

Verified 00c3a254

Adding static_loss_scale to unfused optimizer (#546)

samyam committed 5 years ago

Verified bcd56f97

Support non-tensor state in checkpoint (#548)

tjruwase committed 5 years ago

Verified 6021b702

Fix unbalanced gradients bug in ZeRO-2 gradient accumulation (#545)

tjruwase committed 5 years ago

Verified 0178e6cc

bump version 0.3.4

jeffra committed 5 years ago

6b28bc5d

Discover variables for NCCL backend on AML without mpi4py (#542)

awan-10 committed 5 years ago

Verified 1b45917c

Older