DeepSpeed (research) developer at @microsoft
-
Microsoft
- Seattle, WA
- https://rasley.io
- @jeffra45
Highlights
- Pro
Block or Report
Block or report jeffra
Report abuse
Contact GitHub support about this user’s behavior. Learn more about reporting abuse.
Report abusePinned
-
microsoft/DeepSpeed Public
DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
-
-
microsoft/Megatron-DeepSpeed Public
Forked from NVIDIA/Megatron-LM
Ongoing research training transformer language models at scale, including: BERT & GPT-2
1,048 contributions in the last year
Less
More
Activity overview
Contributed to
microsoft/DeepSpeed,
microsoft/DeepSpeed-MII,
jeffra/sandbox
and 13 other
repositories
Contribution activity
June 2023
Created 7 commits in 1 repository
Opened 1 pull request in 1 repository
microsoft/DeepSpeed
1
closed
Reviewed 8 pull requests in 1 repository
microsoft/DeepSpeed
8 pull requests
- remove UtilsBuilder load, use torch (un)flatten ops
- Documentation for DeepSpeed Accelerator Abstraction Interface
- FP8 unittest for H100
- Update Dockerfile with newer cuda and torch.
- Single Node is using unreferenced pdsh kill cmd while terminating
- [Bugfix][CPU] Remove C++ version in CPU OpBuilder
- Increase tensor creator coverage
- [MoE] remove deepcopy to avoid incompatible torch.jit.script modules.
10
contributions
in private repositories
Jun 2 – Jun 15





