Comment by jtap

Same, but with vision models. Unfortunately, I might be at my limit locally. I have three models that I'm using to find and identify objects in pictures. The largest dataset and model now takes about 8 hours per epoch on my Mac M4 with 16G memory.