Open bjkomer opened 3 years ago
I get the same error if I change it to out = tf.keras.layers.LSTM(50)(inp)
so this might be a TensorFlow-Keras issue instead?
Yea, looks like it is actually an issue on their end, I opened one on there: https://github.com/tensorflow/tensorflow/issues/48235 I also noticed if you don't include metrics it works fine.
Minimal example to reproduce the issue: This was tested with tf version 2.4.1 and keras_lmu version 0.3.1
Note that if
model.fit
is not called, the model will be loaded fine.The error that occurs is: