Dataeval

📝 **Miscellaneous**
- `e7a284de` - Update dataset split unit tests
- `f8731a44` - Add initial logging framework and unit test
- `771dc1d1` - Add conda tests to pipeline
- `2d9fd55a` - Update RTD yaml to use uv for installation
- `0ab99a7f` - Initial prototyping of underspecification tests

📝 **Miscellaneous**
- `102664de` - Remove tensorflow from project
- `e782dad1` - Refactor OutputMetadata and clean up set_metadata decorator
- `80aae3a6` - Just use KSOutput as a MappingOutput instance instead of extracting the dict attribute it no longer has.
- `b738e01f` - Allow docker cmds within dev container
- `16839b46` - Add MappingOutput class
- `e2cfda94` - Made metadata_tools/ks_compare compatible with new KSOutput class.

🌟 **Feature Release**
- `73c1e1be` - Implement PyTorch AutoEncoder based OOD detector

Adds initial PyTorch based Autoencoder OOD detector available when installed with the `torch` extra.

🛠️ **Improvements and Enhancements**
- `70794b5f` - Moved discretization of metadata out of bias functions

📝 **Miscellaneous**
- `4d94e602` - Added test assertions for how_to notebooks
- `7723e242` - Introduce Pytorch OOD detector, with its new training procedure, into OOD howto notebook.
- `f5ac4bdd` - Added new KSOutput class and adapted tests and other functions accordingly
- `3a01a81a` - Introduce new Pytorch OOD detection into prototype metadata demo notebooks.
- `dc155554` - Fix torch gmm functions and enable tests
- `a715c1ef` - Adjust docs to incorporate new metadata function
- `0719bad0` - Update dependencies to remove hdbscan

👾 **Fixes**
- `cac3e2b8` - Fixes drift with pre-processing and shuffles MNIST by default

📝 **Miscellaneous**
- `bacbd0e7` - Use build script specifically for docs
- `0a87e912` - docker build for docs only
- `671b60a5` - Prototype function to infer whether a 1D sample is continuous or discrete
- `d0b8004a` - Use explicit re namespace for compile, search, sub, and MULTILINE
- `502ca2df` - Change to nox for automation test scripts
- `5b46ebea` - Add new bias functional tests and set groundwork for rediscretization

🌟 **Feature Release**
- `e055acf0` - Metadata utility function to merge, extend and flatten metadata
- `95b28ae1` - Adjust bias plotting functions to return figure

📝 **Miscellaneous**
- `532f92a2` - Minimum spanning tree and Clusterer are rewritten using numba for large code speed up
- `7377e012` - Switch jobs to use uv and tox natively
- `7af75016` - Add lazyloading for tensorflow modules

🛠️ **Improvements and Enhancements**
- `ba52ef2e` - Refactor away _internal module

📝 **Miscellaneous**
- `6e55451c` - Integration of distribution compare and OOD MI metadata tools (continued)
- `e4f82173` - Streamlined tests
- `ac8fe3ee` - Fix type mismatch on training AEGMM
- `6289c7d0` - Add plotting helper functions to diversity and balance
- `14d0cfd4` - Integration of low-level metadata drift/OOD exploration functions