Repository navigation
Support loading legacy-serialized probabilistic layer configs - #71
Merged
Merged
Conversation
dnerini
added a commit
that referenced
this pull request
Sep 28, 2026
Resolve conflicts in favour of the torch line: #71's TFP from_config shims are obsolete here (they remain on the v0.x line). Keep main's py3.12 support and dev security pins, bump version to 1.0.0, install CPU-only torch in CI and relock. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
26 of 30 tasks
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Distribution layers saved before
@register_keras_serializable()was introduced serialized theirLambda/DistributionLambdafunction as Python bytecode (function,arguments,make_distribution_fn, etc.), which isn't portable across Python versions and causes old saved models to fail loading on newer Python/TF. This adds afrom_configclassmethod to each probabilistic layer that strips these legacy, no-longer-accepted keys before reconstructing the layer, restoring backward compatibility without changing serialization for newly saved models.