Version 0.7.8

This version supports PyTorch 2.1 to 2.4

New feature

• hausdorff distance / Add "max" point reduction for chamfer distance 44702fd
• limited support for AMD GPUs. Pulsar not included

Improvements

• allow matrix_to_quaternion onnx export 7edaee7
• remove dependency on fvcore 4df110b

Fixes

• obj material loading when some unspecified names but not all 9acdd67
• Fix: setting FrameData.crop_bbox_xywh for backwards compatibility d0d0e02