-
Notifications
You must be signed in to change notification settings - Fork 6.3k
[docs] Parallel loading of shards #12135
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
Conversation
HuggingFaceDocBuilderDev
commented
Aug 13, 2025
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.
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.
Do we want to talk a bit about device_map? The motivation behind passing device_map essentially comes from this PR: #11904. I have also tried providing a bit of motivation behind adding it at the pipeline-level here: #12122 (comment).
I didn't mention #11904 since it doesn't appear to require any action on a user's behalf and it just works in the background. It's a cool design/implementation detail though and I think it'd be pretty interesting to maybe do a blog post about optimizations like this.
@stevhliu it comes to fruition when we initialize the model directly on the accelerator device through device_map instead of doing to("cuda"). So, it could be important to mention I guess.
Docs to go along with #12028.
I didn't mention #11904 since it doesn't appear to require any action on a user's behalf and it just works in the background. It's a cool design/implementation detail though and I think it'd be pretty interesting to maybe do a blog post about optimizations like this.