Try using 32-bit version of the model. I had the same issue, and was using 16-bit model but when switched to 32-bit the model loading time was faster on 18 Pro
Topic:
Machine Learning & AI
SubTopic:
Core ML