Open liaopeiyuan opened 4 years ago
Thanks for this issue. I will look into this. We are currently in the process of providing jit support for all PyTorch modules, so please stay tuned!
Do you potentially have a direction to where it may go wrong? I will be working closely on integrating torch_geometric with tvm the following months, so I may be able to help with some of the issues.
š Bug
There seems to be a type error when tracing a Graph-UNet with Torch JIT.
RuntimeError: unsupported output type: Tensor?
To Reproduce
Steps to reproduce the behavior:
Expected behavior
Correctly returns a value of type
torch.jit.TopLevelTracedModule
Environment
Additional context
I'm working on compilation and deployment of models written with torch_geometric to be deployed on Jetson Nano with tvm, which requires JIT tracing first.