Thanks, it seems to me the best to way to fix this issue is not to require users to supply the correct input names, since this is an impl detail of torch and we can figure out the names in `from_pytorch`.
Instead, users should just supply a list of input shapes in correct order. We call `get_graph_input_names()` after we run `_run_jit_passes` as you proposed, and zip names with a list of shape to get `input_shapes`. This way, users don't have to bother with input names anymore and we can also remove `_check_input_names`. The only downside is that the API would deviate from other frontends. Personally I don't think that is an issue and we should an make API that most makes sense for Torch. Since Torch frontend is still new, we can fix it now. What do you think? cc @alexwong @zhiics @pyjhzwh If we all agree to change the API (`input_shapes` would become a list of shapes, not a dict as in other frontend), we can proceed to fix it. --- [Visit Topic](https://discuss.tvm.ai/t/pytorch-frontend-graph-input-names-can-change-using-loaded-torchscript/6055/2) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.ai/email/unsubscribe/a40ade0a31644c5f61ce1b2b4f2cad5e716dc43864beba7add66cf78e00be565).
