t-vi commented on pull request #7023: URL: https://github.com/apache/tvm/pull/7023#issuecomment-738647194
So what I didn't do yet is move all utility functions that use the state (default_dtype and prelude) into the class. We could add that or defer. Also, I wondered if you're attached to excepting the quantized ops in `report_missing_conversions` and then adding them to the `convert_map` later in the `from_pytorch` function. If one moves the addition to `convert_map` up, one can drop the special casing in the `report_missing_conversions`. (Also, I'm not sure if one wouldn't add the quantized ops all along, but I did not study the design enough here.) ---------------------------------------------------------------- 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]
