huifengguo edited a comment on issue #17728: Poor performance when we use mx.symbol.Variable(..,..,stype="row-sparse") to define the variable URL: https://github.com/apache/incubator-mxnet/issues/17728#issuecomment-593251480 > you can set lazy_update=False for Adam to force disable sparse update. This will give you the same result as dense weight/gradient Do you mean set lazy_update=False for Adam is equivalent to forcing disable sparse update? I always think above two setting are independent with each other... BTW, my purpose is training the model with the sparse embedding table, so the feature of sparse grad is necessary. :(
---------------------------------------------------------------- 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
