Skip to content

ZeRO 3 Offload#834

Merged
samyam merged 17 commits intomasterfrom
staging-zero3-release
Mar 8, 2021
Merged

ZeRO 3 Offload#834
samyam merged 17 commits intomasterfrom
staging-zero3-release

Conversation

@samyam
Copy link
Copy Markdown
Contributor

@samyam samyam commented Mar 8, 2021

ZeRO 3 Offload
---> Supports all three stages of ZeRO
---> Allows for optionally offloading parameters, gradients and optimizer states

jeffra and others added 13 commits March 3, 2021 06:02
Co-authored-by: Samyam <samyamr@microsoft.com>
Co-authored-by: Jeff Rasley <jerasley@microsoft.com>
Co-authored-by: Samyam Rajbhandari <samyamr@microsoft.com>
Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com>
Co-authored-by: Shaden Smith <Shaden.Smith@microsoft.com>
Co-authored-by: Shaden Smith <ShadenTSmith@gmail.com>
Co-authored-by: eltonzheng <eltonz@microsoft.com>
* fp16 Z3 API update and bugfix

* revert debug change
* trying out ZeRO-3 race condition fix

* CUDA sync instead of stream

* reduction stream sync

* remove commented code
Co-authored-by: Jeff Rasley <jerasley@microsoft.com>
Co-authored-by: Jeff Rasley <jerasley@microsoft.com>
@samyam samyam merged commit 599258f into master Mar 8, 2021
@RezaYazdaniAminabadi
Copy link
Copy Markdown
Contributor

Congrats @samyam, @ShadenSmith, @jeffra, @tjruwase, and @eltonzheng 💯

@RezaYazdaniAminabadi
Copy link
Copy Markdown
Contributor

and, all DeepSpeed :)

@mrwyattii mrwyattii deleted the staging-zero3-release branch July 7, 2023 02:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants