docs(RESULTS): static-int8 full-set gate — 4.6241, quality-clean, +8% speed - #216
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The static-activation artifact scored 4.6241 full-set (n=1200) vs the shipped dynamic int8's 4.5701 — +0.054pp, inside the artifact-drift band (the 2.0 artifact measured 5.0329 vs checkpoint 5.08) — while carrying +8% CPU decode speed from the framing matrix. Static int8 clears quality and wins speed. The re-export of all quantized artifacts (calibration stage in modal_export, index-v6, golden regen) is recorded as a release decision with a release bar: the gate re-runs with --out so the delta ships with a paired-bootstrap CI.