[layers] Compatible with Keras !!!

Author: zsdonghao

docs/modules/layers.rst

@@ -318,6 +318,7 @@ Layer list
    TileLayer
 
    SlimNetsLayer
+   KerasLayer
 
    PReluLayer
 
@@ -610,6 +611,14 @@ see `Slim-model <https://github.com/tensorflow/models/tree/master/slim#Install>`
 
 .. autoclass:: SlimNetsLayer
 
+Connect Keras
+------------------
+
+Yes ! Keras models can be connected into TensorLayer! see `tutorial_keras.py <https://github.com/zsdonghao/tensorlayer/blob/master/example/tutorial_keras.py>`_ .
+
+.. autoclass:: KerasLayer
+
+
 Parametric activation layer
 ---------------------------
 

example/tutorial_keras.py

@@ -0,0 +1,75 @@
+#! /usr/bin/python
+# -*- coding: utf8 -*-
+
+import numpy as np
+import tensorflow as tf
+import tensorlayer as tl
+import time
+from keras import backend as K
+from keras.layers import *
+from tensorlayer.layers import *
+
+X_train, y_train, X_val, y_val, X_test, y_test = \
+                tl.files.load_mnist_dataset(shape=(-1, 784))
+
+sess = tf.InteractiveSession()
+
+batch_size = 128
+x = tf.placeholder(tf.float32, shape=[None, 784])
+y_ = tf.placeholder(tf.int64, shape=[None,])
+
+def keras_block(x, is_train=True):
+    x = Dropout(0.8)(x)
+    x = Dense(800, activation='relu')(x)
+    x = Dropout(0.5)(x)
+    x = Dense(800, activation='relu')(x)
+    x = Dropout(0.5)(x)
+    logits = Dense(10, activation='linear')(x)
+    return logits
+
+network = InputLayer(x, name='input')
+network = KerasLayer(network, keras_layer=keras_block, name='keras')
+
+y = network.outputs
+network.print_params(False)
+network.print_layers()
+
+cost = tl.cost.cross_entropy(y, y_, 'cost')
+correct_prediction = tf.equal(tf.argmax(y, 1), y_)
+acc = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
+
+n_epoch = 200
+learning_rate = 0.0001
+
+train_params = network.all_params
+train_op = tf.train.AdamOptimizer(learning_rate, beta1=0.9, beta2=0.999,
+    epsilon=1e-08, use_locking=False).minimize(cost, var_list=train_params)
+
+tl.layers.initialize_global_variables(sess)
+
+for epoch in range(n_epoch):
+    start_time = time.time()
+    ## Training
+    for X_train_a, y_train_a in tl.iterate.minibatches(
+                                X_train, y_train, batch_size, shuffle=True):
+        _, _ = sess.run([cost, train_op], feed_dict={x: X_train_a, y_: y_train_a,
+                                K.learning_phase(): 1})
+
+    print("Epoch %d of %d took %fs" % (epoch + 1, n_epoch, time.time() - start_time))
+    ## Evaluation
+    train_loss, train_acc, n_batch = 0, 0, 0
+    for X_train_a, y_train_a in tl.iterate.minibatches(
+                            X_train, y_train, batch_size, shuffle=False):
+        err, ac = sess.run([cost, acc], feed_dict={x: X_train_a, y_: y_train_a,
+                                K.learning_phase(): 0})
+        train_loss += err; train_acc += ac; n_batch += 1
+    print("   train loss: %f" % (train_loss/ n_batch))
+    print("   train acc: %f" % (train_acc/ n_batch))
+    val_loss, val_acc, n_batch = 0, 0, 0
+    for X_val_a, y_val_a in tl.iterate.minibatches(
+                                X_val, y_val, batch_size, shuffle=False):
+        err, ac = sess.run([cost, acc], feed_dict={x: X_val_a, y_: y_val_a,
+                                K.learning_phase(): 0})
+        val_loss += err; val_acc += ac; n_batch += 1
+    print("   val loss: %f" % (val_loss/ n_batch))
+    print("   val acc: %f" % (val_acc/ n_batch))

tensorlayer/layers.py

@@ -4413,9 +4413,11 @@ def __init__(
         layer = None,
         slim_layer = None,
         slim_args = {},
-        name ='InceptionV3',
+        name ='tfslim_layer',
     ):
         Layer.__init__(self, name=name)
+        assert slim_layer is not None
+        assert slim_args is not None
         self.inputs = layer.outputs
         print("  [TL] SlimNetsLayer %s: %s" % (self.name, slim_layer.__name__))
 
@@ -4444,6 +4446,43 @@ def __init__(
         self.all_layers.extend( slim_layers )
         self.all_params.extend( slim_variables )
 
+## Keras layer
+class KerasLayer(Layer):
+    """
+    The :class:`KerasLayer` class can be used to merge all Keras layers into
+    TensorLayer. Example can be found here `tutorial_keras.py <https://github.com/zsdonghao/tensorlayer/blob/master/example/tutorial_keras.py>`_
+
+    Parameters
+    ----------
+    layer : a list of :class:`Layer` instances
+        The `Layer` class feeding into this layer.
+    keras_layer : a keras network function
+    keras_args : dictionary
+        The arguments for the keras model.
+    name : a string or None
+        An optional name to attach to this layer.
+    """
+    def __init__(
+        self,
+        layer = None,
+        keras_layer = None,
+        keras_args = {},
+        name ='keras_layer',
+    ):
+        Layer.__init__(self, name=name)
+        assert layer is not None
+        assert keras_layer is not None
+        self.inputs = layer.outputs
+        print("  [TL] KerasLayer %s: %s" % (self.name, keras_layer))
+        with tf.variable_scope(name) as vs:
+            self.outputs = keras_layer(self.inputs, **keras_args)
+            variables = tf.get_collection(TF_GRAPHKEYS_VARIABLES, scope=vs.name)
+        self.all_layers = list(layer.all_layers)
+        self.all_params = list(layer.all_params)
+        self.all_drop = dict(layer.all_drop)
+        self.all_layers.extend( [self.outputs] )
+        self.all_params.extend( variables )
+
 ## Special activation
 class PReluLayer(Layer):
     """