Keras: Cannot load_model

[abhisheksoni27]

Thank you!

• [ ] Check that you are up-to-date with the master branch of Keras. You can update with:
  pip install git+git://github.com/keras-team/keras.git --upgrade --no-deps

• [x] If running on TensorFlow, check that you are up-to-date with the latest version. The installation instructions can be found here.

• [ ] If running on Theano, check that you are up-to-date with the master branch of Theano. You can update with:
  pip install git+git://github.com/Theano/Theano.git --upgrade --no-deps

• [x] Provide a link to a GitHub Gist of a Python script that can reproduce your issue (or just copy the script here if it is short).

I am using Google Colab to train a CNN and then save the entire model to a .h5 file. The code is available here: CNN-Colab

The model gets saved but when I later try to load it back, I get the following error:

TypeError: float() argument must be a string or a number, not 'dict'

The entire Output log is here: CNN - Colab - Error

[yanboliang]

@abhisheksoni27 I think you hit this bug, it has been fixed at keras-team/keras, but you are using tf.keras. Actually I have sent a fix to tf.keras as well, but not merged until now.

[ymodak]

Closing this issue since the fixed has been merged. Feel free to reopen the issue if the problem still persists. Thanks!

[psoulos]

Is there a way to apply the fix as a patch to Tensorflow 1.12 installed in Anaconda?

Edit:
For those looking to fix this issue, follow these steps.

1. Open a python interpreter
2. import tensorflow
3. Type tensorflow.__file__ to find the location of your tensorflow install (call this $TENSORFLOW_DIR)
4. Navigate to your tensorflow install directory
5. Edit $TENSORFLOW_DIR/python/keras/layers/advanced_activations.py
6. In the stack trace where you saw the TypeError: float() argument must be a string or a number message, a few lines up there should be a trace from advanced_activations.py.
7. Edit the line in advanced_activations.py to use a try/except block

try: self.SOME_VALUE = K.cast_to_floatx(SOME_VALUE) except TypeError: self.SOME_VALUE = K.cast_to_floatx(SOME_VALUE['value'])