wzh99 commented on PR #11681: URL: https://github.com/apache/tvm/pull/11681#issuecomment-1153073465
@ganler Thanks for your suggestion. However, I am sorry that I cannot completely agree that this test should be placed in [test_type_infer.py](https://github.com/apache/tvm/blob/main/tests/python/relay/test_type_infer.py). The reason is that the tests in this file are concerned with the language features of Relay itself (tuple, ref, let-binding, ADT, etc.) There are some test cases about operators such as `test_add_broadcast_op`, but I think they are mainly used to test whether the type inference in Relay can support different forms of type relations, instead of whether the type relations of a specific operator is correct. I still think that the test case should be placed in [test_op_level2.py](https://github.com/apache/tvm/blob/main/tests/python/relay/test_op_level2.py), because this is the place for testing specific type relations of `nn.conv3d`. I can add the following lines at the end of `test_conv3d_infer_type`: ``` # Infer with groups x = relay.var("x", relay.TensorType((1, 16, 224, 224, 224), "float32")) w = relay.var("w", relay.TensorType((4, 4, 1, 1, 1), "float32")) y = relay.nn.conv3d(x, w, groups=4, kernel_size=(1, 1, 1), channels=4) yy = run_infer_type(y) assert yy.checked_type == relay.TensorType((1, 4, 224, 224, 224), "float32") ``` I am new to contributing to TVM and my understanding may not be thorough enough. Please point out if I make any mistakes. -- 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]
