Skip to content

[Native] Replay sample sets without decoding their tokens - #24950

Draft
metascroy wants to merge 1 commit into
offgraph-kv-mg-gguffrom
offgraph-kv-verify-without-tokenizer
Draft

metascroy wants to merge 1 commit into
offgraph-kv-mg-gguffrom
offgraph-kv-verify-without-tokenizer

Conversation

@metascroy

@metascroy metascroy commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Stack

Off-graph KV cache for the native ggml engine, through serving Muse Glimmer on it. Review bottom-up; each PR builds and passes its tests on its own.

Summary

--verify_samples streamed each generated token through --tokenizer, which only decodes Llama 3 vocabularies and threw on any other model's token IDs, though replay compares logits and never needs text. generate() now takes the tokenizer as optional and streams text only when given one; sample replay passes none.

Test plan

  • Muse Glimmer packages, whose token IDs exceed Llama 3's vocabulary, replay their golden samples.
  • Existing packages still verify, and normal generation still streams text.

Authored with Claude Code.

@pytorch-bot

pytorch-bot Bot commented Nov 12, 2026 •

Copy link
Copy Markdown

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24950

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

❌ 137 New Failures, 7 Unrelated Failures, 10 Unclassified Failures

As of commit b5c3fb1 with merge base c7f7931 (image):

NEW FAILURES - The following jobs have failed:

UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

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

@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Nov 12, 2026
--verify_samples streamed each generated token through --tokenizer, which
only reads text for Llama 3 vocabularies and threw on any other model's
token IDs, though the replay compares logits and never needs text.
generate() now takes the tokenizer as optional and streams text only when
given one; sample replay passes none.

This branch had an error being deployed

1 failed deployment
cadence — b5c3fb16 Deployed Nov 14, 2026 by metascroy via hifi5s-build / hifi5s #38327
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant