leezu commented on issue #16926: autograd.is_training() does not work on hybridized networks. URL: https://github.com/apache/incubator-mxnet/issues/16926#issuecomment-559380807 Hybridization traces the input during the first forward pass to the hybridized network. Afterwards it will re-use the computational graph established during the initial trace. To solve your issue, simply create two instances of your network. Both should share parameters and both can be hybridized. Just use one of them for training, the other one for validation. You're very welcome to submit a PR that clarifies the documentation.
---------------------------------------------------------------- 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
