-
Notifications
You must be signed in to change notification settings - Fork 3.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
DeepSpeed Infinity Update #7234
Conversation
Codecov Report
@@ Coverage Diff @@
## master #7234 +/- ##
======================================
- Coverage 92% 91% -0%
======================================
Files 203 203
Lines 13137 13157 +20
======================================
- Hits 12023 12005 -18
- Misses 1114 1152 +38 |
Should've made this draft since we can't merge this yet |
After some debugging, fix for the latest DeepSpeed CI here: deepspeedai/DeepSpeed#1026 We'll need a new version to be added to support this, so might be a while before we can merge this PR. I can refer in the documentation to the PR and get this working, so people can test it out :) |
Co-authored-by: Carlos Mocholí <[email protected]>
# Conflicts: # .azure-pipelines/gpu-tests.yml # pytorch_lightning/plugins/training_type/deepspeed.py
# Conflicts: # pytorch_lightning/plugins/training_type/deepspeed.py
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM !
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Awesome!
Remember to go through the checklist in the description
What does this PR do?
Updates the API to the latest changes in DeepSpeed + exposes parameters required for using NVMe Offloading.
This hard requires the latest DeepSpeed, however considering that this is still experimental and changes here are future proofing, I think this is fine.
Before submitting
PR review
Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:
Did you have fun?
Make sure you had fun coding 🙃