Actually, this can happen in the body of the function, but not here because the inputs actually come from a function signature. You can print `traced_module.code` to witness the translation (that is from where I tracked down the function reproducing the non-processed names). Another place where you can the argument names them directly and programmatically is in the schema of the traced module's forward method: `[a.name for a in traced_module.forward.schema.arguments]`.
I haven't fully investigated what it would take to make PyTorch present the signature in a way retrievable by `inspect.signature`(it currently isn't available), that might be the best way to present it. That said, people appear to prefer the current API with its requirement to pass names and shapes regardless of whether they are already provided by the module, so I guess it'll have to stay that way. I just thought that it would help to analyze what is going on in the JIT to make informed decisions about how to convert models. Best regards Thomas --- [Visit Topic](https://discuss.tvm.ai/t/pytorch-frontend-graph-input-names-can-change-using-loaded-torchscript/6055/20) 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/7e53506a2ad7a5856062e8b64c1b36d9aa77dfa2172ddf7ea54bf8a840ee5089).
