fix(export): derive the ONNX output contract from num_out - #360
Merged
Merged
Conversation
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #357. The export sidecar always described the output as a single BCE logit, so a multiclass checkpoint shipped a contract telling consumers to apply a sigmoid to a
[batch, num_out]softmax. The output block now comes from the checkpoint: binary exports keep the old block byte for byte; multiclass exports get[batch, num_out], a softmax convention, class order fromlabel_map, and the calibration method whencalibration.jsonsits beside the checkpoint. Export refuses to write a contract whose width disagrees with the graph, or whoselabel_mapdoes not index exactly0..num_out-1.A multiclass checkpoint with no
label_mapstill exports, withclasses: nulland a warning, since training can produce that shape and the column order is still correct.https://claude.ai/code/session_01JDZiRc4dgzaA7nYFijxGbz