feat(perf): add strict reported-zero prefill ladder and sizing guide - #21
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #15. Stacked after the explicit thinking/deadline changes.
Adds a separate
prefill-ladder-v1.json: nominal 2k/8k/32k/128k repeated-text inputs, one-token output cap, one warmup plus three measured trials per size, concurrency 1, explicit enable-thinking false, and strictreported-prefix-zeroevidence. Existing sparkDash examples are unchanged.Documents worked 128k and hypothetical 256k admission sizing, including the difference between response_bytes and request size, JSON escaping, fill bytes and reservation copies. Names are approximate until provider-reported token usage is observed. The 256k row is sizing guidance, not a shipped or tested case.
Verified: independent SIX/HELP review, exact-head fmt/Clippy/workspace tests/release build, and an actual CLI loopback run of the unmodified 16-wave example. Fixtures prove missing/nonzero cache usage remains ineligible and reported zero qualifies. These are synthetic protocol checks, not tokenizer measurements.
Unverified: actual 128k model prefill, provider template compliance and real prefill speeds. No live inference requests were authorized or made. This resolves the requested runnable example and sizing documentation; it does not claim live 128k qualification.