Quadra

Updated

- Keep only opencv-python-headless as a dependency for quadra to avoid conflicts with the non-headless version

Updated

- Terminate quadra with error if it's not possible to export an ONNX model with automatic mixed precision instead of falling back to full precision

Fixed

- Fix default quadra installation requiring extra dependencies incorrectly
- Fix matplotlib using interactive backend by default
- Fix documentation errors

Updated

- Update anomalib version, improve release workflow

Updated

- Update dependencies to support publishing Quadra to PyPI

Updated

- Improve safe batch size computation for sklearn based classification tasks

Fixed

- Fix wrong dtype used when evaluating finetuning or anomaly models trained in fp16 precision