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

Adds a new end-to-end test for Mistral 7b #903

Merged
merged 1 commit into from
Sep 23, 2024

Conversation

shralex
Copy link
Collaborator

@shralex shralex commented Sep 19, 2024

The changes include:

  • A notebook in scratch_pad that runs the model from HF, for sample inputs and generates logits, which it stores in a file.
  • A test that downloads the HF checkpoint, converts it to MaxText-compatible format, and uses it to do one forward pass on a sample input.
  • Compares the logits obtained in both steps above to make sure they're equal.

The old test, which asserted on the generated text being equal, rather than logits, is removed.

Copy link
Collaborator

@khatwanimohit khatwanimohit left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

@@ -0,0 +1,42 @@
#!/bin/bash

# This file runs on daily basis and demonstrates:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: mention where we run this test. v4-8 in this case.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess this file should be split into 2 parts like the rest, but that is unrelated to fix here so okay to leave for now

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It was previously split, and merged into one by Mohit's request, since this is a small model.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ah makes sense, sounds good!

- A notebook in scratch_pad that runs the model from HF, for sample inputs
and generates logits, which it stores in a file.
- A test that downloads the HF checkpoint, converts it to MaxText-compatible format,
and uses it to do one forward pass on a sample input.
- Compares the logits obtained in both steps above to make sure they're equal.
The old test, which asserted on the generated text being equal, rather than logits, is removed.
@copybara-service copybara-service bot merged commit eb1e772 into main Sep 23, 2024
13 checks passed
@copybara-service copybara-service bot deleted the shralex_mistral_test branch September 23, 2024 15:53
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants