masahi commented on issue #5167: [RELAY][QNN]Add support for QNN Div operator URL: https://github.com/apache/incubator-tvm/pull/5167#issuecomment-609003422 @siju-samuel What frontend are you using? Torch frontend can already convert and run quantized mobilnet v3. I didn't have a need for QNN div op. For hswish, Torch uses `quantized::mul_scalar` op, which is a scaling by 1/6. For scaling by positive scalar, I didn't have to do anything. Just changing the "interpretation" of the input quantized tensor is all I needed. See https://github.com/apache/incubator-tvm/blob/03cbf78e3e8cc910c5a9c95cd3fcafa19959644f/python/tvm/relay/frontend/qnn_torch.py#L697-L701
---------------------------------------------------------------- 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] With regards, Apache Git Services
