Skip to content
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

Inference with the checkpoint #586

Open
mathmax12 opened this issue Sep 23, 2024 · 3 comments
Open

Inference with the checkpoint #586

mathmax12 opened this issue Sep 23, 2024 · 3 comments
Labels
enhancement New feature or request

Comments

@mathmax12
Copy link

Though there are some related tickets, it would be great to see the steps to use the raw checkpoints for inference. https://github.com/pytorch/torchtitan/blob/main/docs/checkpoint.md
This helps to verify the training works well.

@tianyu-l tianyu-l added the enhancement New feature or request label Sep 24, 2024
@tianyu-l
Copy link
Contributor

torchtitan is place to do training. For inference maybe you can try torchchat?

@jaysonfrancis
Copy link

jaysonfrancis commented Oct 19, 2024

@tianyu-l Some basic generation may help discover/test issues in checkpoint conversion, tokenization, validation before continued pre training etc.. I am working on adapting a minimal version from torchtune to support basic train-time evals during checkpointing

@tianyu-l
Copy link
Contributor

@jaysonfrancis Covering the very basic generation/inference sounds good, as long as the code can be kept simple. Please feel free to submit a PR, we can help review.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

3 participants