- Refactor DatasetRNN to be subclassable for different target types and introduce subclasses for the existing target types. DatasetRNN can no longer be used directly.
0.1.4 - 2026-01-22
- Modify default behavior of DatasetRNN to use random batching, and allow passing a random number generator for reproducibility.
- Modify DatasetRNN to return a dict instead of a tuple. This paves the way for generalizations with fields other than just inputs and targets.
0.1.3 - 2026-01-15
- Fix a bug which prevented some two_armed_bandit agents and environments from working correctly to generate synthetic datasets
- Add a new loss for jointly training on continuous and categorical targets. The intended use-case is NeuroDisRNN.
- Improve typing for rnn params
- Fix plotting functions for NeuroDisRNN
0.1.2 - 2025-11-17
- Validate input types to DatasetRNN
- Improve docstrings in rnn_utils
- Update code for plotting two-armed bandit session data and add examples to the notebook
0.1.1 - 2025-11-10
- Initial Release on PyPi
Previous version was not released on PyPI, so bumping up the version to trigger a release and act as a baseline for future releases.
- Initial release