Brandon N Benton
Brandon N Benton
As of scikit-learn version 1.2 the `sparse` parameter for `OneHotEncoder` has been renamed to `sparse_output`. Update name and pin scikit-learn >= 1.2 or pin < 1.2?
Ok, here we go... `sup3r/preprocessing` was previously just data handlers and batch handlers, essentially. Now we have `Loaders`, `Extracters`, `Derivers`, `Cachers` which are composed in `sup3r.preprocessing.data_handlers.factory` to build objects similar...
This adds the option to skip chunks in the forward pass routine if all coordinates covered by that chunk are masked. The mask is provided as an additional variable through...
Added recursive layer initialization in `HiddenLayers` and `Sup3rObsModel` which can be built from nested hidden layers in a config, and also ingest multiple obs features / exo features.
How would we feel about adding an issue to deprecate the random forest stuff? I don't think we ever use this. I think we built this originally when testing rf...
Python 3.12+ requires Tensorflow 2.16+, which requires Keras 3. Keras 3 implemented a complete rewrite enabling "multi-framework machine learning", using any of JAX, TensorFlow, and PyTorch and moving between them...
The commit history is a mess but the cumulative effect is almost entirely in sup3r/preprocessing/rasterizers/exo.py and sup3r/models/with_obs.py. Lots of tiny edits re: refactoring, adding doc strings, etc.
**Why this feature is necessary:** Resolving this would enable us to use larger batches and possibly train with data that hasn't been moved to locally mounted SSDs. **A possible solution...
Small fix for idw_fill with 4D models. Deprecating random forest code