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

document temperature and system prompt sensitivity #1626

Merged
merged 4 commits into from
Feb 8, 2025

Conversation

prabhu
Copy link
Collaborator

@prabhu prabhu commented Feb 8, 2025

  • cdx1 is very sensitive to temperature and the system prompt. This makes sense since the unsloth fine-tune has switched the architecture of phi-4 to llama and is therefore consistent with observations like these. Our approach to using unsloth/phi-4 is further validated by the fact, microsoft has now updated phi-4 configs using the unsloth patches.
  • speculative decoding is supported with mlx!

Signed-off-by: Prabhu Subramanian <prabhu@appthreat.com>
Signed-off-by: Prabhu Subramanian <prabhu@appthreat.com>
Signed-off-by: Prabhu Subramanian <prabhu@appthreat.com>
Signed-off-by: Prabhu Subramanian <prabhu@appthreat.com>
@prabhu prabhu marked this pull request as ready for review February 8, 2025 16:58
@prabhu prabhu merged commit 95bd5b2 into master Feb 8, 2025
16 checks passed
@prabhu prabhu deleted the feature/cdx1-eval-notes branch February 8, 2025 16:58
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.

1 participant