Inference 1.7.3: Faster RF-DETR on Mac

Inference 1.7.3 is now available. RF-DETR now runs through Core ML on Apple silicon, about 3x faster than on CPU in our tests with RF-DETR Nano on an M4 Max.

The release also adds the Embedding Model block, which pulls feature vectors from a classifier you already trained (ResNet, ViT, or DINOv3) so you can feed them to Cosine Similarity for novelty detection or similarity search without retraining.