AmigoCDT edited a comment on issue #14229: About the weighted softmax, the forward result is too small URL: https://github.com/apache/incubator-mxnet/issues/14229#issuecomment-470778540 @piyushghai yes, I use another similar customOP (Focal loss), it works well. So the trouble is in my code. Two choices : 1. gradient: `gradient passing` or `gradient formulation`. 2. I lose something in my script. If there is no fault at my gradient. The main error get this: `mx.metric.EvalMetric`, Focal loss CustomOP also rewrite this class. But I do not know what this helps. I will find at mxnet doc.
---------------------------------------------------------------- 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
