Releases: lume-science/lume-torch
v2.0.1
What's Changed
- Update README.md by @roussel-ryan in #132
- Enable loading NN ensembles saved in JIT by @RemiLehe in #134
- Fix non-PSD failing test by @pluflou in #135
Note: these changes might be incompatible with multitask GP implementations that used botorch versions <0.15.
Full Changelog: v2.0.0...v2.0.1
v2.0.0
Breaking Changes
torchandbotorchare now part of the base requirementsregister_to_mlflowno longer takes aninputexample as an argument
New Features
- TorchModel input and output transformers now accept a Callable type
- allow transformers to be callables by @kathryn-baker in #128
Bug Fixes
Full Changelog: v1.9.3...v2.0.0
v1.9.3
v1.9.2
v1.9.1
v1.9.0
v1.8.1
v1.8.0
What's Changed
- Support torch.nn.Linear transformers by @t-bz in #101
- Add input and output validation by @pluflou in #104
- Add validation by @t-bz in #102
- Add global random seed for deterministic tests by @allrob23 in #95
- Update python versions by @pluflou in #105
- Make minimum Python version be 3.10
- Improve variable validation by @pluflou in #106
- adding save_jit option to save scripted jit module by @RevathiJambunathan in #103
- Fix bug in default validation by @pluflou in #110
- add jit filename to yaml and load jit if available by @pluflou in #111
- Merging updates from main by @pluflou in #113
- Add tests and docs by @pluflou in #107
- Add linting using ruff/pre-commit hooks
- Update docs to reflect changes above
- Adjust codebase to support latest Botorch
Note that the variable classes have changes, and now only ScalarVariable is used. Other subclasses: InputVariable, ScalarInputVariable etc were deprecated. Changes to attribute names for ScalarVariable class have also been updated, notably: variable_type is now variable_class (specified as a lume_model.variables class), default is now default_value (required for TorchModel inputs), and all input default values need to be within the value_range provided for all TorchModel inputs.
Examples of the new API can be found in the README and the example notebooks.
New Contributors
- @allrob23 made their first contribution in #95
- @RevathiJambunathan made their first contribution in #103
Full Changelog: v1.7.1...v1.8.0
v1.7.1
LUME-Model v1.7.0
What's Changed
- Discontinue support for Python 3.7 & 3.8 and pin pydantic version by @t-bz in #84
- Register transformers and add option to pass a model to PyTorchModel constructor by @t-bz in #85
- Make LUMEBaseModel pydantic and change in-/output variables to lists by @t-bz in #86
- Clean up and restructure model subpackages by @t-bz in #87
- Pydantic v2 by @roussel-ryan in #88
- Allow relative paths in YAML files and update docs by @t-bz in #89
New Contributors
- @roussel-ryan made their first contribution in #88
Full Changelog: v1.6.0...v1.7.0