-
Notifications
You must be signed in to change notification settings - Fork 376
Update sagemaker docs structure #1645
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
Mar 24, 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.
this file is not showing up as a page anymore--I can include it as a section called "Examples" if helpful
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.
Yes I think we should keep this page for now. There's not only a list of notebook examples but also a useful doc on the inference toolkit.
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.
Not sure what makes sense for the title of the page. There are not only examples but also API specs, so I wouldn't call it "Examples". maybe the best is to keep "Reference" for now?
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.
Sure, we can keep it
Hello Arjuna, thanks for the PR!
A few comments:
I would keep all the containers info in a single page if possible, same as here: https://huggingface.co/docs/google-cloud/en/containers/available.
This way, our customers don't need to know the name of our libraries to browse through the containers
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.
We should add the import in the code snippet
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.
The util doesn't work for Pytorch training DLC, no?
Hello Arjuna, thanks for the PR! A few comments: I would keep all the containers info in a single page if possible, same as here: https://huggingface.co/docs/google-cloud/en/containers/available. This way, our customers don't need to know the name of our libraries to browse through the containers
Sounds reasonable. We can keep it in one file for now and then restructure later. That being said, if the headings in the same file are still "TGI", "Transformers", etc, customers will still need to know the library names to know what to use, similarly to if these are the titles of subheadings.
Is there a better way to denote the differences that might be more intuitive?
Sounds reasonable. We can keep it in one file for now and then restructure later. That being said, if the headings in the same file are still "TGI", "Transformers", etc, customers will still need to know the library names to know what to use, similarly to if these are the titles of subheadings. Is there a better way to denote the differences that might be more intuitive?
Yes I think the list of containers deserve a short intro / how to section where you explain the rational on how to pick the right container for your usecase.
For a basic intro, I'd say there are two main things to explain:
- whether you want to do inference or training (Pytorch Training DLC vs others)
- knowing if your model is supported by an available inference toolkit DLC which should yield better perf than the Pytorch Inference DLC (generic purpose). A small decision tree could be very cool here.
We could have a more complex version of the decision tree in an "advanced section". It could take into account more factors like quantization, target hardware etc.. This doc doesn't exist but becomes more necessary now TGI is multibackend.
Btw, I think the name we use are Pytorch Inference DLC and Pytorch Training DLC over Transformers DLC. I don't know if it's a good name and we can change it, as long as we stay consistent.
Update sagemaker docs structure to remove references to old images, as well as structure in a way where automated PRs will update tables