Support tf.keras

[yu-iskw]

Issue

#73

Overview

If an environment has tf.keras, tf.keras is preferentially used to pure keras.

Possible issue

If an user wan to use pure keras but the environment is not a tensorflow backend, an error might occur.

[philipperemy]

I have this error when I run it:

from tensorflow.keras.engine.topology import Layer
ModuleNotFoundError: No module named 'tensorflow.keras.engine'

[philipperemy]

Package              Version Location
-------------------- ------- -------------------------------
absl-py              0.8.0
astor                0.8.0
gast                 0.3.1
google-pasta         0.1.7
grpcio               1.23.0
h5py                 2.10.0
Keras-Applications   1.0.8
Keras-Preprocessing  1.1.0
Markdown             3.1.1
numpy                1.17.2
pip                  19.2.3
protobuf             3.9.1
setuptools           41.2.0
six                  1.12.0
tensorboard          1.14.0
tensorflow           1.14.0
tensorflow-estimator 1.14.0
termcolor            1.1.0
Werkzeug             0.15.6
wheel                0.33.6
wrapt                1.11.2

[philipperemy]

python -c 'from tensorflow.keras.engine.topology import Layer'

[yu-iskw]

The namespace was wrong. I modified it. Can you take a look at it?

[mimxrt]

ImportError: cannot import name 'Input' from 'tensorflow.keras.models' (unknown location)

change:

from tensorflow.keras.models import Input, Model

to

from tensorflow.keras import Input, Model

[mimxrt]

Another error:

/ws/miniconda3/lib/python3.7/site-packages/tcn/tcn.py in __call__(self, inputs, training)
    148                                              dropout_rate=self.dropout_rate,
    149                                              kernel_initializer=self.kernel_initializer,
--> 150                                              use_batch_norm=self.use_batch_norm)
    151                 skip_connections.append(skip_out)
    152         if self.use_skip_connections:

/ws/miniconda3/lib/python3.7/site-packages/tcn/tcn.py in residual_block(x, training, dilation_rate, nb_filters, kernel_size, padding, activation, dropout_rate, kernel_initializer, use_batch_norm)
     55     # 1x1 conv to match the shapes (channel dimension).
     56     prev_x = Conv1D(nb_filters, 1, padding='same')(prev_x)
---> 57     res_x = keras.layers.add([prev_x, x])
     58     res_x = Activation(activation)(res_x)
     59     return res_x, x

NameError: name 'keras' is not defined

I fixed it by removing the keras. prefix in two occasions and changing

import tensorflow.keras.layers

to

import tensorflow.keras.layers as layers

[yu-iskw]

@mimxrt Thank you for making sure of it. I have modified some including the tasks.

[philipperemy]

See #94