I'm having issues trying to get YOLOv11 working with PyTorch MPS on apple silicon. Keep getting this C++ ops error and can't figure out what's wrong. Here's my setup: M3 Pro MacBook Python 3.9 torch 2.6.0 torchvision 0.21.0 torchaudio 2.6.0 ultralytics 8.3.70 Simple test code that's failing: from ultralytics import YOLO model = YOLO("yolov8n.pt") results = model.train(data="coco8.yaml", epochs=100, imgsz=640, device="mps") This is the error I keep getting: RuntimeError: Couldn't load custom C++ ops. This can happen if your PyTorch and torchvision versions are incompatible, or if you had errors while compiling torchvision from source. For further information on the compatible versions, check https://github.com/pytorch/vision#installation for the compatibility matrix. MPS is available (checked with torch.backends.mps.is_available()). I've tried the obvious stuff - reinstalling the packages, updating to latest versions, etc. No luck. Also tried using nightly versions, and following the ultralytics/Apple docs but they are very limited: https://developer.apple.com/metal/pytorch/ https://docs.ultralytics.com/modes/train/#can-i-train-yolo11-models-on-apple-silicon-chips Anyone run into this before? Really want to use my M3's GPU if possible rather than falling back to CPU. Thanks! Continue reading...
