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

Improve Test Vector Generation #8

Merged
merged 4 commits into from
Nov 4, 2024

Conversation

Xeratec
Copy link
Collaborator

@Xeratec Xeratec commented Oct 30, 2024

This PR improves the test vector generation and ensure that we have non-constant output vectors. Furthermore, it fixes the too small bit width of the weights.

Added

  • Print and raise warnings for problematic vectors
  • Print and check the error of the partial streaming softmax relative to the integer softmax

Changed

  • Reduce calculated bit width of AV (This assumes that the attention is distributed among different values and not on one token)

Fixed

  • Fix wrong bitwidth of weights (the -1 is already handled in the random_shuffled_tensor function)

- This assumes that the attention is distributed among different values and not on one token
@Xeratec Xeratec requested a review from gamzeisl October 30, 2024 22:13
Copy link
Collaborator

@gamzeisl gamzeisl left a comment

Choose a reason for hiding this comment

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

LGTM

PyITA/ITA.py Show resolved Hide resolved
@gamzeisl gamzeisl merged commit 774c356 into pulp-platform:main Nov 4, 2024
3 checks passed
@Xeratec Xeratec deleted the pr/improve_vector_generation branch November 4, 2024 13:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants