Skip to content

MLPro-BF: Fix for Pickling Compatibility with PyTorch 2.4+

Choose a tag to compare

@steveyuwono steveyuwono released this 12 Mar 08:40
· 885 commits to main since this release

This release addresses an issue with object serialization in PyTorch versions later than 2.4.0. To ensure compatibility, we have set the recurse=True parameter in the dill library, allowing objects to be broken down into smaller components for proper pickling. This fix has been tested with MLPro-Int-SB3 and is now merged into the main repository. With this update, serialization in MLPro should work across different PyTorch versions.

What's Changed

MLPro-BF - Basic Functions

  • Bug SB3: Unit Test Failed due to Pickling Serialized Objects by @steveyuwono in #1165

Full Changelog: v1.9.7...v1.9.8