Repository navigation
fix(model): checkpoints always carry fresh bias_int - #25
Merged
Merged
Conversation
… save Stale calibration-time biases cost 0.95 dB on the M3 val gate (first eval +5.81 dB, refreshed +6.76 dB). Checkpoints now always carry bias_int matching the weights they store; shifts stay frozen as trained.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 M3 gate story in code form: bias_int computed at epoch-2 calibration went stale over 147 further epochs of bias training, costing 0.95 dB on the integer path (+5.81 first eval vs +6.76 after refresh). train() now refreshes bias_int from the current FP biases before every checkpoint save; shifts remain frozen as trained. Two new tests: the refresh formula unit test and a checkpoint-freshness integration test.