Quadra

Fixed

- Fix deepcopy removing model signature wrapper from the model for classification

Fixed

- Fix pytorch model mlflow upload. Not supported.

Updated

- Update torch to 2.1.2 with CUDA 12 support
- Update pytorch lightning to 2.1.*

Changed

- Refactor hydra plugins to use optional dev groups intend of extras to avoid dragging local packages around in external installations
- Refactor extra dev dependencies to use poetry groups instead
- Improve trainer configs to avoid wrong overrides when calling different trainer overrides

Fix

- Fix ONNX import call for the utilities when ONNX is not installed.

Added

- Add upload_models to core config

Refactored

- infer_signature_torch_model refactored to infer_signature_model

Added

- Add support for half precision training and inference for sklearn based tasks
- Add gradcam export for sklearn training