mbrookhart opened a new pull request #8241: URL: https://github.com/apache/tvm/pull/8241
Adding more ops to support QAT BERT. I also refactored the tests for easier extensions. I marked this as WIP because the result isn't matching tflite, I need to do some more comparisons over the next day or two to isolate the problem. I'm opening the PR now because supporting ops with multiple outputs required refactoring how I handle quantization specific types. Bringing it into the make ir namespace allows me to do what I need to do, but it also opens up a question of where or not this could be more used more generally in other places for quantization? cc @jroesch @masahi @anijain2305 @jwfromm -- 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]
