[Fix] Fix Method to Obtain Prefix Token ID

[anzr299]

Summary

Fix prefix_token_id to return the BOS token ID instead of the EOT token ID.

EagerEvalWrapper.prefix_token_id was incorrectly returning the EOT token ID.
Since lm-eval prepends prefix_token_id to every evaluation sequence, this caused
Llama 3's <|end_of_text|> (token 128001) to be used instead of <|begin_of_text|>
(token 128000), resulting in higher perplexity scores.

Llama 3 8B Wikitext PPL before fix : 9.18
Llama 3 8B Wikitext PPL after fix : 7.793

The result after the fix matches the expected perplexity when evaluating
the same model directly via HuggingFace.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18317

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 6 Awaiting Approval, 1 New Failure

As of commit 55fd4d1 with merge base 1925873 ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

• Build Cadence (gh)
• Build documentation (gh)
• Build Presets (gh)
• Lint (gh)
• periodic (gh)
• pull (gh)

NEW FAILURE - The following job has failed:

• Copilot code review / Cleanup artifacts (gh)
  Process completed with exit code 1.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[Copilot]

Pull request overview

Fixes Llama evaluation behavior by returning a proper prefix token id (BOS when available) instead of incorrectly defaulting to the end-of-text/end-of-sequence token, aligning perplexity results with Hugging Face’s eval flow.

Changes:

• Update prefix_token_id to prefer tokenizer.bos_id when present.
• Preserve prior fallback behavior by using eot_token_id when BOS is unavailable.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.