microsoft/DeepSpeed

Pull Requests Commits

Fixed allgather mem_factor

Quentin-Anthony committed 3 years ago

45e6b350

Fix ds_bench, fix args, fix dtypes, fix bw_unit

Quentin-Anthony committed 3 years ago

c6370044

No longer package ds_bench with setup.py

Quentin-Anthony committed 3 years ago

c3ec062f

Add ds_bench to setup and README

Quentin-Anthony committed 3 years ago

240df2ca

Added ds_bench, abstracted out args, error handling

Quentin-Anthony committed 3 years ago

babab4f7

Remove unnecessary DS imports and double headers

Quentin-Anthony committed 3 years ago

917da1ce

Merge branch 'qanthony/comms-bench' of https://github.com/microsoft/DeepSpeed into qanthony/comms-bench

Quentin-Anthony committed 3 years ago

682b4aa3

Move benchmarks

Quentin-Anthony committed 3 years ago

34caa84e

Delete all_reduce_v2.py

Quentin-Anthony committed 3 years ago

Verified d8a5f3f9

Delete runall.py

Quentin-Anthony committed 3 years ago

Verified feaa1499

Make debug alltoall-specific

Quentin-Anthony committed 3 years ago

0edfbaff

Quentin-Anthony committed 3 years ago

655f1131

Add pt2pt and README. Make mem_factor an arg.

Quentin-Anthony committed 3 years ago

31f96941

Change single runs to automatically use max message size

Quentin-Anthony committed 3 years ago

b154d359

Add WIP benchmarks

Quentin-Anthony committed 3 years ago

15ca89e7

test credentials

Quentin-Anthony committed 3 years ago

243858b8

Add v1 allreduce benchmark

Quentin-Anthony committed 3 years ago

cd2fecf7

[CI] force upgrade HF dependencies & output py env (#2015)

jeffra committed 3 years ago

Verified e6f444ae

update CODEOWNERS (#2017)

conglongli committed 3 years ago

Verified 117c9cdf

Relax assertion to allow Megatron-DeepSpeed MoE to use ZeRO-1 (#2007)

Quentin-Anthony committed 3 years ago

Verified 25b2fc29

DeepSpeed comm backend v1 (#1985)

awan-10 committed 3 years ago

Verified 36ad3119

[docs] add new build badges to landing page (#1998)

jeffra committed 3 years ago

Verified 828ab718

Propagate max errorcode to deepspeed when using PDSH launcher (#1994)

jerrymannil committed 3 years ago

Verified d0eae5ad

[bug] Add user-defined launcher args for MPI launcher (#1933)

mrwyattii committed 3 years ago

Verified 3678ee17

Add torch-latest and torch-nightly CI workflows (#1990)

mrwyattii committed 3 years ago

Verified 7fc30650

fix typo when getting kernel dim in conv calculation (#1989)

cli99 committed 3 years ago

Verified 6719b46b

added temp fix for our CI jobs (#1988)

mrwyattii committed 3 years ago

Verified b6f2a560

Add unit test for various model families and inference tasks (#1981)

mrwyattii committed 3 years ago

Verified 5d3c6718

small fix for the HF Bert models (#1984)

RezaYazdaniAminabadi committed 3 years ago

Verified 0ebd81df

add 530b paper (#1979)

jeffra committed 3 years ago

Verified 3da84185

Older