leandron commented on pull request #9599: URL: https://github.com/apache/tvm/pull/9599#issuecomment-980783160
> Thanks @leandron for bringing it up! The error message is only supposed to appear when user doesn’t supply arch information, and it could not be detected in our local environment- which means either the local environment is not set up properly, or it’s cross compilation without a clear target arch. Either case looks suspicious, and thus we think it’s necessary to provide error message explicitly. > > Would you like to elaborate when the warning looks unnecessarily verbose? Is it in the CI environment? Sure, if you see the example I provide in the initial message, it is expected that some machines won’t have a Rocm and/or cuda version - they are optional dependencies anyway. For those machines, the user will always see those warning messages, even when, for example, they would only want to target CPU with LLVM. This is what I meant by it being overly verbose, from the user point of view. Perhaps we could make it more specific on when those messages are displayed, so that users see them when they are relevant in the context. I don’t have any proposal to solve that, but moving them to debug will make both the API and tvmc usage more clean. -- 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]
