sxjscience commented on issue #16716: [Numpy] Fix collect_params().zero_grad() in gluon numpy interface URL: https://github.com/apache/incubator-mxnet/pull/16716#issuecomment-551243605 > scare me, frankly. Doing numpy-like fully imperative execution has no chance of being actually fast and leaning into that direction will not make MXNet any better. Gluon seems like a good compromise - sure, do imperative if you just debug but hybridize when you are actually serious about getting something deployable. And the fix proposed in this PR: if this is NumPy-like array then do slow thing (especially since as I understand it numpy style will be promoted going forward) is super bad practice. Leaning towards numpy does not mean to throw away the mixing of symbolic and imperative. It's more about the front-end interface to be numpy-like. Also, doing **slow** things is not always bad. Sometimes it will improve the readability and make the package more researcher-friendly.
---------------------------------------------------------------- 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: us...@infra.apache.org With regards, Apache Git Services
