NCCL-based 1-bit Adam + Code Refactor for Comm. Backends (#594)
a6dba72a
Merge branch 'master' into staging-1bit-nccl-v2
78400850
Revert "Merge branch 'master' into staging-1bit-nccl-v2"
6dbdd985
Revert "Revert "Merge branch 'master' into staging-1bit-nccl-v2""
9712f10c
Merge branch 'master' into staging-1bit-nccl-v2
4d1f4f01
comm optimization + 1-bit lamb
d8a23c98
Saving/debugging commit.
89e19362
finalizing 1-bit lamb
a1bbf781
finalizing 1-bit lamb
db0ca769
add momentum mask and chkpt handling for 1-bit adam
07deab84
Merge remote-tracking branch 'origin/staging-1bit-nccl-v2' into stagiā¦
625f475f
Cleanup and modify nccl test to be runnable with deepspeed launcher.
d55fddb5
Merge branch 'master' into staging-1bit-nccl-v2
5b1cacb7
Fix format.
8cbc212b
fix formatting again.
ff8c871a
make test runnable without mpi4py
c17041f3
Add dist.alltoall and dist.allgather instead of custom functions.
5e01a30e
remove debug prints.
97a55577
formatting and renaming
e3e1e39b
renaming
d5b9dcc8
renaming
3d66a8a2
add unit test, fix existing tests
b042467e
Merge branch 'master' into staging-1bit-adam-v2
ab3521d1
skip unit test when torch < 1.8
9fa5166c
revert 1-bit lamb
65d7ec59
flatten momentum when dimension is more than 1
8376a404
add warning message for 1-bit adam under fp32
6a19f296
improve version check
819043da
add fp32 test
a6943be9
Merge remote-tracking branch 'origin' into staging-1bit-adam-v2
2042b299
1-bit adam doc
66a8c930
fix file name
fb329a9c
doc fix
0b3c1d76
torch 1.8 is released
0bffa9b1
doc fix
294c2d6f
fix tests
003981a4
Merge branch 'master' into staging-1bit-adam-v2
3f42b3a4
Merge branch 'master' into staging-1bit-adam-v2
bbd61436
update news
877f8d71
Merge branch 'master' into staging-1bit-adam-v2
f861465c
add doc for momentum mask
2ed029e3
Merge branch 'staging-1bit-adam-v2' of github.com:microsoft/DeepSpeedā¦
c6e7cf79
fix checkpoing handling, add unit test
3b53c90a
checkpoint handling doc
42407294
doc final cleanup
968a53fd
Merge branch 'master' into staging-1bit-adam-v2
bbec3007
Merge branch 'master' into staging-1bit-adam-v2
e28a99e9
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
jeffra
commented
on 2021-03-16
bump dates
1221aec8
Merge branch 'staging-1bit-adam-v2' of github.com:microsoft/DeepSpeedā¦
535b5bad
jeffra
requested changes
on 2021-03-16
update tests
8cfd2b78
url change
38ff08a9
doc fix
de036564
fix test
5957bce9
doc update
ef51ac69
Merge branch 'master' into staging-1bit-adam-v2
7c08b34d
jeffra
approved these changes
on 2021-03-16
conglongli
merged
68c8481b
into master 5 years ago
conglongli
deleted the staging-1bit-adam-v2 branch 4 years ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub