i have problem with torch.compile()

[Sariskiattitha]

I've come across the PyTorch 2.0 lecture by MrDBourke, and it's really interesting to learn how to make the model train very fast.

Then, I have two questions:

1.) I use torch.compile(model) in the forward step, and it returns an error. Has anyone else come across this problem?

This is the error.

2.) Does anyone know how to allocate more Apple GPU memory to PyTorch? Because in the documentation, torch.mps seems not to have that command, or I really can't find it.

Sincerely, thanks!!

[shisirkha]

Hmm, you might go see the actual documentation for this