Kevin-Li-2025 opened a new pull request, #19818: URL: https://github.com/apache/tvm/pull/19818
### Description Fixes #19574. ONNX `BatchNormalization` defaults `training_mode` to integer `0`, while Relax `nn.batch_norm` expects a boolean `training` attribute. Passing the integer through lets the imported op carry a non-bool training attr and can route default ONNX BatchNormalization through training-mode semantics instead of inference/running-stat semantics. This converts the ONNX attribute to `bool` before constructing `relax.op.nn.batch_norm`, and adds an importer-level regression test that verifies default ONNX BatchNormalization is imported with `training is False`. ### Tests - `python3 -m py_compile python/tvm/relax/frontend/onnx/onnx_frontend.py tests/python/relax/test_frontend_onnx.py` - Static regression check confirming the bool conversion and new frontend test - `git diff --check` The full TVM Python test cannot run in this local checkout because `tvm_ffi` is not built/installed. -- 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]
