Does the "torch.cuda.is_available()" option affect kernel crash?

[Victini00]

While running demo_part3.ipynb, cell 3 crashes due to kernel issue.

I think this is due to torch.cuda.is_available() option. I found that torch.cuda.is_available() is False on cell 2.

Am I right? If yes, should I make torch.. available() true by setting torch and cuda environment? my cuda version is 12.1.

Thanks for reading it.