Hi
I have only recently started working on ML on my Mac M1 Pro, previously I was working on a Windows platform. I am having difficulties getting my machine set up right so that its ready for the super fast training I was hoping for when I got it. Please help me with this and let me know if and where I am going wrong.
So, I tried a custom dataset training using Yolov8 model. I want to train for a 100 epochs. Now the same dataset and hyperparameters take about 2.5 hours on a T4 GPU on Google Colab, whereas I was only at around 60 epochs after 24 hours on my M1 pro.
I have home brew, miniconda, pytorch nightly for mac installed and set the device to mps when training the YOLO model. I feel that this is reaaallly slow. What should I be doing right?
Thank you Lakshmi