search

vlawhern / arl-eegmodels

This is the Army Research Laboratory (ARL) EEGModels Project: A Collection of Convolutional Neural Network (CNN) models for EEG signal classification, using Keras and Tensorflow
Other
1.17k stars 287 forks source link

Keras >= 2.1.5 compatibility #2

Closed heilerich closed 6 years ago

heilerich commented 6 years ago

The required Keras version is specified as >=2.1.3, but the module is only compatible with versions 2.1.3 and 2.1.4.

Keras has moved the DepthwiseConv2D class to another module. To be compatible with more recent Keras versions the import path has to be changed. If you want to keep compatibility with Keras 2.1.3/4 I would suggest a conditional import with a try/catch statement.

Greetings Felix

vlawhern commented 6 years ago

Sorry for the late response... been rather busy as of late.

A colleague of mine notified me of this a while back. Thanks for the reminder. I have a version of the code that uses Tensorflow only, opting to use the internal Tensorflow Keras library as opposed to a separate Keras install as it is done now. I'll also update the dependency to the newest Tensorflow once I can get to it.

So I'm gonna let this stay open for now until I can push the new version.

vlawhern commented 6 years ago

Closing this issue, as we now use Keras as provided inside of Tensorflow (as opposed to a separate Keras install). Also updated Tensorflow requirement to 1.8.0 and made changes to README.md accordingly.