Skip to content

[Examples] Add Eagle and Eagle3 Conversion Scripts #56

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

Draft
wants to merge 7 commits into
base: main
Choose a base branch
from

Conversation

dsikka
Copy link
Collaborator

@dsikka dsikka commented Jul 16, 2025

Summary

Unconverted Models Trained Models

Eagle Models:

  • nm-testing/Eagle_Speculator_Llama_3_1_8B_TTT (converted with and without fusion-bias)

Eagle3 Models:

  • nm-testing/Speculator-Qwen3-8B-Eagle3 (Qwen verifier/target model)
  • nm-testing/SpeculatorLlama3-1-8B-Eagle3 (Llama verifier/target model)

rahul-tuli
rahul-tuli previously approved these changes Jul 16, 2025
Copy link

github-actions bot commented Jul 16, 2025

📦 Build Artifacts Available
The build artifacts (`.whl` and `.tar.gz`) have been successfully generated and are available for download: https://github.com/neuralmagic/speculators/actions/runs/16474250609/artifacts/3597997443.
They will be retained for up to 30 days.
Commit: ae845e8

@dsikka dsikka requested a review from MeganEFlynn July 16, 2025 20:25
shanjiaz
shanjiaz previously approved these changes Jul 22, 2025
@dsikka dsikka marked this pull request as ready for review July 23, 2025 14:58
@dsikka
Copy link
Collaborator Author

dsikka commented Jul 23, 2025

@MeganEFlynn For Eagle3 - I can remove the examples to not show both with and without norm_before_residual - what would we want for Llama and Qwen?

@dsikka dsikka dismissed stale reviews from shanjiaz and rahul-tuli via ae845e8 July 31, 2025 18:47
@dsikka dsikka marked this pull request as draft August 6, 2025 12:11
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.

3 participants