Hi there,
It seems to me that TorchScript is a subset of Pythran, or at least that we could enlarge Pythran to make it a superset of PyTorch. If that’s the case, it would be interesting to have both compiler share the same AST optimization engine.
Pythran optimizations have been designed to be relatively independent from the target, there’s probably room for cooperation there.
Does that look like a good idea?