PawelGlomski-Intel edited a comment on pull request #20292: URL: https://github.com/apache/incubator-mxnet/pull/20292#issuecomment-923009647
@szha The changes I made to the native convolution introduce a slight (~8% on my machine) drop in performance in its backward pass. Is the performance of native implementation crucial (backward pass in particular)? Native 2D deconvolution seems to be slower after my changes too, but as it wasn't working properly (#11203), I am not sure if I can compare them 1:1. Also, are you aware why #20101 introduced [this](https://github.com/barry-jin/incubator-mxnet/blob/4006219d83d151a48707b2ad292d3579e52c1561/python/mxnet/ndarray/numpy_extension/_op.py#L616) assert? -- 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]
