fix keras examples

examples/conv_filter_visualization.py

@@ -41,7 +41,7 @@ def deprocess_image(x):
     return x
 
 # build the VGG16 network with ImageNet weights
-model = vgg16.VGG16(weights='imagenet', include_top=False)
+model = vgg16.VGG16(weights='imagenet', include_top=False, input_shape=(img_width, img_height, 3))
 print('Model loaded.')
 
 model.summary()

examples/conv_lstm.py

@@ -9,6 +9,10 @@
 from keras.layers.normalization import BatchNormalization
 import numpy as np
 import pylab as plt
+from keras import backend as K
+
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: ConvLSTM2D Layer is not supported yet.")
 
 # We create a layer which take as input movies of shape
 # (n_frames, width, height, channels) and returns a movie

examples/deep_dream.py

@@ -19,6 +19,8 @@
 from keras.applications import inception_v3
 from keras import backend as K
 
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: Symbolic Gradients is not supported yet.")
 parser = argparse.ArgumentParser(description='Deep Dreams with Keras.')
 parser.add_argument('base_image_path', metavar='base', type=str,
                     help='Path to the image to transform.')

examples/imdb_lstm.py

@@ -19,6 +19,7 @@
 from keras.layers import Dense, Embedding
 from keras.layers import LSTM
 from keras.datasets import imdb
+from keras import backend as K
 
 max_features = 20000
 maxlen = 80  # cut texts after this number of words (among top max_features most common words)
@@ -37,8 +38,15 @@
 
 print('Build model...')
 model = Sequential()
-model.add(Embedding(max_features, 128))
-model.add(LSTM(128, dropout=0.2, recurrent_dropout=0.2))
+
+# MXNet backend does not support dropout in LSTM and cannot automatically infer shape
+if K.backend() == 'mxnet':
+    # specifying input_length and removed dropout params
+    model.add(Embedding(max_features, 128, input_length=maxlen))
+    model.add(LSTM(128, unroll=True))
+else:
+    model.add(Embedding(max_features, 128))
+    model.add(LSTM(128, dropout=0.2, recurrent_dropout=0.2))
 model.add(Dense(1, activation='sigmoid'))
 
 # try using different optimizers and different optimizer configs

examples/lstm_seq2seq.py

@@ -122,14 +122,18 @@
             decoder_target_data[i, t - 1, target_token_index[char]] = 1.
 
 # Define an input sequence and process it.
-encoder_inputs = Input(shape=(None, num_encoder_tokens))
+# MXNet backend RNN required input shape, for TensorFlow backend, you can provide the shape as:
+# encoder_inputs = Input(shape=(None, num_encoder_tokens))
+encoder_inputs = Input(shape=(max_encoder_seq_length, num_encoder_tokens))
 encoder = LSTM(latent_dim, return_state=True)
 encoder_outputs, state_h, state_c = encoder(encoder_inputs)
 # We discard `encoder_outputs` and only keep the states.
 encoder_states = [state_h, state_c]
 
 # Set up the decoder, using `encoder_states` as initial state.
-decoder_inputs = Input(shape=(None, num_decoder_tokens))
+# MXNet backend RNN required input shape, for TensorFlow backend, you can provide the shape as:
+# decoder_inputs = Input(shape=(None, num_decoder_tokens))
+decoder_inputs = Input(shape=(max_decoder_seq_length, num_decoder_tokens))
 # We set up our decoder to return full output sequences,
 # and to return internal states as well. We don't use the
 # return states in the training model, but we will use them in inference.

examples/mnist_denoising_autoencoder.py

@@ -36,8 +36,14 @@
 (x_train, _), (x_test, _) = mnist.load_data()
 
 image_size = x_train.shape[1]
-x_train = np.reshape(x_train, [-1, image_size, image_size, 1])
-x_test = np.reshape(x_test, [-1, image_size, image_size, 1])
+if K.image_data_format() == 'channels_first':
+    x_train = np.reshape(x_train, [-1, 1, image_size, image_size])
+    x_test = np.reshape(x_test, [-1, 1, image_size, image_size])
+    input_shape = (1, image_size, image_size)
+else:
+    x_train = np.reshape(x_train, [-1, image_size, image_size, 1])
+    x_test = np.reshape(x_test, [-1, image_size, image_size, 1])
+    input_shape = (image_size, image_size, 1)
 x_train = x_train.astype('float32') / 255
 x_test = x_test.astype('float32') / 255
 
@@ -52,7 +58,6 @@
 x_test_noisy = np.clip(x_test_noisy, 0., 1.)
 
 # Network parameters
-input_shape = (image_size, image_size, 1)
 batch_size = 128
 kernel_size = 3
 latent_dim = 16

examples/mnist_swwae.py

@@ -94,7 +94,7 @@ def getwhere(x):
     y_prepool, y_postpool = x
     return K.gradients(K.sum(y_postpool), y_prepool)
 
-if K.backend() == 'tensorflow':
+if K.backend() != 'theano':
     raise RuntimeError('This example can only run with the '
                        'Theano backend for the time being, '
                        'because it requires taking the gradient '

examples/neural_doodle.py

@@ -55,6 +55,9 @@
 from keras.preprocessing.image import load_img, img_to_array
 from keras.applications import vgg19
 
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: Symbolic Gradients is not supported yet.")
+
 # Command line arguments
 parser = argparse.ArgumentParser(description='Keras neural doodle example')
 parser.add_argument('--nlabels', type=int,

examples/neural_style_transfer.py

@@ -60,6 +60,9 @@
 from keras.applications import vgg19
 from keras import backend as K
 
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: Symbolic Gradients is not supported yet.")
+
 parser = argparse.ArgumentParser(description='Neural style transfer with Keras.')
 parser.add_argument('base_image_path', metavar='base', type=str,
                     help='Path to the image to transform.')

examples/pretrained_word_embeddings.py

@@ -38,7 +38,8 @@
 print('Indexing word vectors.')
 
 embeddings_index = {}
-with open(os.path.join(GLOVE_DIR, 'glove.6B.100d.txt')) as f:
+# change encoding to utf-8 for glove.6B dataset
+with open(os.path.join(GLOVE_DIR, 'glove.6B.100d.txt'), encoding='utf-8') as f:
     for line in f:
         values = line.split()
         word = values[0]

examples/variational_autoencoder.py

@@ -53,7 +53,8 @@ def sampling(args):
 xent_loss = original_dim * metrics.binary_crossentropy(x, x_decoded_mean)
 kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
 vae_loss = K.mean(xent_loss + kl_loss)
-
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: Custom loss is not supported yet.")
 vae.add_loss(vae_loss)
 vae.compile(optimizer='rmsprop')
 vae.summary()

examples/variational_autoencoder_deconv.py

@@ -120,6 +120,8 @@ def sampling(args):
     K.flatten(x_decoded_mean_squash))
 kl_loss = - 0.5 * K.sum(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
 vae_loss = K.mean(xent_loss + kl_loss)
+if K.backend() == 'mxnet':
+    raise NotImplementedError("MXNet Backend: Custom loss is not supported yet.")
 vae.add_loss(vae_loss)
 
 vae.compile(optimizer='rmsprop')

keras/backend/mxnet_backend.py

@@ -3501,6 +3501,28 @@ def truncated_normal(shape, mean=0.0, stddev=1.0, dtype=None, seed=None):
     return KerasSymbol(sym)
 
 
+# CTC(Connectionist Temporal Classification)
+
+def ctc_batch_cost(y_true, y_pred, input_length, label_length):
+    """Runs CTC loss algorithm on each batch element.
+
+    # Arguments
+        y_true: tensor `(samples, max_string_length)`
+            containing the truth labels.
+        y_pred: tensor `(samples, time_steps, num_categories)`
+            containing the prediction, or output of the softmax.
+        input_length: tensor `(samples, 1)` containing the sequence length for
+            each batch item in `y_pred`.
+        label_length: tensor `(samples, 1)` containing the sequence length for
+            each batch item in `y_true`.
+
+    # Returns
+        Tensor with shape (samples,1) containing the
+            CTC loss of each element.
+    """
+    raise NotImplementedError("MXNet Backend: CTC is not supported yet.")
+
+
 # HIGH ORDER FUNCTIONS
 
 def map_fn(fn, elems, name=None, dtype=None):