[GitHub] anirudh2290 commented on issue #9912: No training happening when CSVIter() is used.
anirudh2290 commented on issue #9912: No training happening when CSVIter() is used. URL: https://github.com/apache/incubator-mxnet/issues/9912#issuecomment-379013025 Opened a new issue for improving documentation: https://github.com/zackchase/mxnet-the-straight-dope/issues/464 . @dimon777 If your issue is resolved can this be closed ? This is an automated message from the Apache Git Service. To respond to the message, please log on 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
[GitHub] anirudh2290 commented on issue #9912: No training happening when CSVIter() is used.
anirudh2290 commented on issue #9912: No training happening when CSVIter() is used. URL: https://github.com/apache/incubator-mxnet/issues/9912#issuecomment-378840196 This doesnt seem like a bug. The reason that you see all those Nans is because you have already iterated through your dataset before the call to evaluate_accuracy. Your first example in mxnet-csviter where you dont see this issue is because evaluate_accuracy doesn't iterate through your dataset. I think the gluon documentation about DataLoader probably was confusing : http://gluon.mxnet.io/chapter04_convolutional-neural-networks/cnn-gluon.html. Maybe it should be called out in the book how to replace DataLoader with non-gluon iterators. You can fix your problem by resetting your iterator after a pass through the data using `iter.reset()` This is an automated message from the Apache Git Service. To respond to the message, please log on 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