ibsidorenko opened a new pull request, #15686: URL: https://github.com/apache/tvm/pull/15686
This commit adds 2 new Relax ops: 1. "relax.op.smooth" (R.smooth): Multiply elements from a tensor by a scale (if it operates like "multiply") or pass input as is (if it operates like "identity"). 2. "relax.annotate.absmax" (R.absmax): Helper op, that does the following transform: abs()->max()->squeeze() for the input tensor. Both of them are used in SmoothQuant TVM passes used in MLC-llm. This PR does not include unit tests for these new ops, since they are single-purpose ops and not widely used. This way we can save time on regression testing. -- 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]
