Closed alexhiggins732 closed 2 years ago
This is unfortunately a known PyTorch+CUDA/GPU issue on Windows, on some configurations (like yours, with only 4GB of RAM) it sometime fails when more than one torch instance is loaded. There are some workarounds though, see the link above. But this is beyond the scope of TorchStudio.
This will be fixed with the next version of PyTorch (1.13) which will be compatible with the latest version of CUDA (11.7) which solves this issue.
Loading the
MNIST Classifier
and attempting to build the model results in the following error.Windows 10 Enterprise 2004 build 19041.1526
Hardware:
Log: