MXNet should support symbolic shape like TF does (I guess), but I have zero knowledge of it, and it's probably broken or out of maintenance or not compatible with deep NumPy api or something :-|
Regarding ur second question, the performance should roughly be the same, since they are both doing reduction operation. (Maybe `npx.logsoftmax` could be faster) -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/apache/incubator-mxnet/issues/19722#issuecomment-754447442
