Aharrypotter commented on PR #19616: URL: https://github.com/apache/tvm/pull/19616#issuecomment-4550414415
Thanks @tlopex, addressed in the latest commit by fixing nested subgraph registration and normalizing empty FlatBuffer input/output vectors. The nested subgraph issue is fixed by registering generated private functions through the shared parent `module_builder`, instead of leaving nested functions inside the temporary subgraph `BlockBuilder`. I also added `test_call_subgraph_nested_call` to cover: ```text main CALL -> subgraph A -> CALL subgraph B ``` This verifies that both generated private functions are present in the final IRModule. I also normalized empty FlatBuffer input/output vectors before iterating or zipping them, including `InputsAsNumpy()`, `OutputsAsNumpy()`, and metadata checks, using an explicit `is not None` fallback path. While touching that code, I switched `_get_subgraph_params` to use `self.get_tensor_type_str(...)` for consistency with the rest of the class. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
