facebookresearch / schedule_free

Schedule-Free Optimization in PyTorch
Apache License 2.0
1.9k stars 64 forks source link

Doesnot work with torch.compile #30

Closed yuanzhi-zhu closed 2 weeks ago

yuanzhi-zhu commented 5 months ago

after adding model = torch.compile(model) in the example/mnist code, I got RuntimeError: Triton Error [CUDA]: device kernel image is invalid

adefazio commented 5 months ago

Thanks for pointing this out, we will investigate

adefazio commented 4 months ago

I added that line and it ran without crashes on my machine, could you provide more information? Full logging output, and what version of PyTorch are you using?