dcslin commented on pull request #733: URL: https://github.com/apache/singa/pull/733#issuecomment-643770454
> @dcslin > In your code, even you use the same train dataset for evaluation, the accuracy will drop from 0.985 (train) to 0.867 (eval) using the same train dataset (test dataset eval accuracy is 0.504 in this case). > > Seems that (i) the inference rountine have some problem with cudnnRNNForwardInference, or (ii) the train model cannot generalize (only can remember the train data but cannot generalize) even though i used m.train() for "inference" on testing set, still the accuracy is low (0.5) ---------------------------------------------------------------- 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]
