All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
Types of changes
Added
for new features.Changed
for changes in existing functionality.Deprecated
for soon-to-be removed features.Removed
for now removed features.Fixed
for any bug fixes.Security
in case of vulnerabilities.
- Parameters without a gradient no longer cause an AttributeError exception. This occurs when a model has a parameter with the
requires_grad
attribute set toFalse
(e.g. the model has a frozen layer). Previously, the exception would be raised when the SM3 class methodstep
is called.
- Updated optimizer to use new PyTorch function signatures as the previous signatures were being deprecated. This means prior releases of PyTorch may not work with this optimizer.
- Updated repository structure for packaging. The code for the optimizer has
been moved to
src/
. The code for testing the optimizer against NumPy has been moved totests/
. - Updated GitHub workflow to create
pip
packaging and test
- Created
PyTorch-SM3
repository and finalized initial code.