I am trying to profile several deep learning models for training and inference using NVIDIA NSight Compute but when I try to profile YOLOv8 either in training or inference mode the profiling is non-terminating and just keeps on happening and I am unable to get a corresponding .ncu-rep file.
Even BERT-base model gets profiled faster than the YOLOv8 model.
For YOLOv8 I have a dataset of 16 images and it infers and trains on the custom dataset in finite time when run without profiling but profiling seems to be non-terminating at all!