TylerADavis opened a new pull request #7134: URL: https://github.com/apache/tvm/pull/7134
This PR builds upon #7128 , adding a test for `is_floating_point()` following masahi's [recommendation](https://github.com/apache/tvm/pull/7128#issuecomment-747862332) to look at `verify_script_model()`. In addition to the test, this PR makes the following changes: * `verify_script_model()` can now pass a dtype to `verify_model_vm()` * `verify_model_vm()` can now generate random inputs for bool and int dtypes. * `verify_model_vm()` includes TVM dtype information in `input_shapes`, providing additional type information to operators such as `is_floating_point()`. These additional changes were made because `verify_model_vm()` does not currently pass type information to `relay.frontend.from_pytorch()`, preventing `is_floating_point()` from working correctly. The changes to random tensor generation were required as `torch.randn()` does not support bool or int dtypes, and my test requires inputs with these dtypes. If it would be helpful, I can split the addition of the test and the changes to testing infrastructure out into two separate PRs. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
