Update IRNN example

fix tensorflow travis config

.travis.yml

@@ -1,18 +1,42 @@
-sudo: false
+sudo: required
+dist: trusty
 language: python
-# Setup anaconda
-before_install:
-  - wget http://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh -O miniconda.sh
-  - chmod +x miniconda.sh
-  - ./miniconda.sh -b
-  - export PATH=/home/travis/miniconda/bin:$PATH
-  - conda update --yes conda
 python:
+  - "2.7"
   - "3.4"
-# command to install dependencies
 install:
-  - conda install --yes python=$TRAVIS_PYTHON_VERSION numpy scipy matplotlib pandas pytest h5py
-  # Coverage packages are on my binstar channel
+  # code below is taken from http://conda.pydata.org/docs/travis.html
+  # We do this conditionally because it saves us some downloading if the
+  # version is the same.
+  - if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then
+      wget https://repo.continuum.io/miniconda/Miniconda-latest-Linux-x86_64.sh -O miniconda.sh;
+    else
+      wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh;
+    fi
+  - bash miniconda.sh -b -p $HOME/miniconda
+  - export PATH="$HOME/miniconda/bin:$PATH"
+  - hash -r
+  - conda config --set always_yes yes --set changeps1 no
+  - conda update -q conda
+  # Useful for debugging any issues with conda
+  - conda info -a
+
+  - conda create -q -n test-environment python=$TRAVIS_PYTHON_VERSION numpy scipy matplotlib pandas pytest h5py
+  - source activate test-environment
+  - pip install pytest-cov python-coveralls
+  - pip install git+git://github.com/Theano/Theano.git
   - python setup.py install
+
+  # install TensorFlow
+  - if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then
+      pip install https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.5.0-cp27-none-linux_x86_64.whl;
+    fi
 # command to run tests
-script: py.test tests/
+script:
+  - PYTHONPATH=$PWD:$PYTHONPATH py.test -v --cov-report term-missing --cov keras tests/
+  - if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then
+      sed -i -e 's/theano/tensorflow/g' ~/.keras/keras.json;
+      PYTHONPATH=$PWD:$PYTHONPATH py.test -v --cov-report term-missing --cov keras tests/;
+    fi
+after_success:
+  - coveralls

LICENSE

@@ -1,6 +1,23 @@
-The MIT License (MIT)
+COPYRIGHT
 
-Copyright (c) 2015 
+All contributions by François Chollet:
+Copyright (c) 2015, François Chollet.
+All rights reserved.
+
+All contributions by Google:
+Copyright (c) 2015, Google, Inc.
+All rights reserved.
+
+All other contributions:
+Copyright (c) 2015, the respective contributors.
+All rights reserved.
+
+Each contributor holds copyright over their respective contributions.
+The project versioning (Git) records all such contribution source information.
+
+LICENSE
+
+The MIT License (MIT)
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -1,8 +1,8 @@
-# Keras: Theano-based Deep Learning library
+# Keras: Deep Learning library for Theano and TensorFlow
 
 ## You have just found Keras.
 
-Keras is a minimalist, highly modular neural network library in the spirit of Torch, written in Python / Theano so as not to have to deal with the dearth of ecosystem in Lua. It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research.
+Keras is a minimalist, highly modular neural networks library, written in Python and capable of running either on top of either [TensorFlow](https://github.com/tensorflow/tensorflow) or [Theano](https://github.com/Theano/Theano). It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research.
 
 Use Keras if you need a deep learning library that:
 - allows for easy and fast prototyping (through total modularity, minimalism, and extensibility).
@@ -12,178 +12,92 @@ Use Keras if you need a deep learning library that:
 
 Read the documentation at [Keras.io](http://keras.io).
 
-Keras is compatible with __Python 2.7-3.4__.
+Keras is compatible with:
+    - __Python 2.7-3.5__ with the Theano backend
+    - __Python 2.7__ with the TensorFlow backend
+
+
+------------------
+
 
 ## Guiding principles
 
 - __Modularity.__ A model is understood as a sequence or a graph of standalone, fully-configurable modules that can be plugged together with as little restrictions as possible. In particular, neural layers, cost functions, optimizers, initialization schemes, activation functions, regularization schemes are all standalone modules that you can combine to create new models.
 
-- __Minimalism.__ Each module should be kept short and simple (<100 lines of code). Every piece of code should be transparent upon first reading. No black magic: it hurts iteration speed and ability to innovate.
+- __Minimalism.__ Each module should be kept short and simple. Every piece of code should be transparent upon first reading. No black magic: it hurts iteration speed and ability to innovate.
 
-- __Easy extensibility.__ New modules are dead simple to add (as new classes/functions), and existing modules provide ample examples. To be able to easily create new modules allows for total expressiveness, making Keras suitable for advanced research.
+- __Easy extensibility.__ New modules are dead simple to add (as new classes and functions), and existing modules provide ample examples. To be able to easily create new modules allows for total expressiveness, making Keras suitable for advanced research.
 
-- __Work with Python__. No separate models configuration files in a declarative format (like in Caffe or PyLearn2). Models are described in Python code, which is compact, easier to debug, and allows for ease of extensibility.
+- __Work with Python__. No separate models configuration files in a declarative format. Models are described in Python code, which is compact, easier to debug, and allows for ease of extensibility.
 
-## Examples
 
-### Multilayer Perceptron (MLP):
+------------------
+
+
+## Getting started: 30 seconds to Keras
+
+The core datastructure of Keras is a __model__, a way to organize layers. There are two types of models: [`Sequential`](/models/#sequential) and [`Graph`](/models/#graph).
+
+Here's the `Sequential` model (a linear pile of layers):
 
 ```python
 from keras.models import Sequential
-from keras.layers.core import Dense, Dropout, Activation
+
+model = Sequential()
+```
+
+Stacking layers is as easy as `.add()`:
+
+```python
+from keras.layers.core import Dense, Activation
+
+model.add(Dense(output_dim=64, input_dim=100, init="glorot_uniform"))
+model.add(Activation("relu"))
+model.add(Dense(output_dim=10, init="glorot_uniform"))
+model.add(Activation("softmax"))
+```
+
+Once your model looks good, configure its learning process with `.compile()`:
+```python
+model.compile(loss='categorical_crossentropy', optimizer='sgd')
+```
+
+If you need to, you can further configure your optimizer. A core principle of Keras is to make things reasonably simple, while allowing the user to be fully in control when they need to (the ultimate control being the easy extensibility of the source code).
+```python
 from keras.optimizers import SGD
-
-model = Sequential()
-model.add(Dense(20, 64, init='uniform'))
-model.add(Activation('tanh'))
-model.add(Dropout(0.5))
-model.add(Dense(64, 64, init='uniform'))
-model.add(Activation('tanh'))
-model.add(Dropout(0.5))
-model.add(Dense(64, 2, init='uniform'))
-model.add(Activation('softmax'))
-
-sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
-model.compile(loss='mean_squared_error', optimizer=sgd)
-
-model.fit(X_train, y_train, nb_epoch=20, batch_size=16)
-score = model.evaluate(X_test, y_test, batch_size=16)
+model.compile(loss='categorical_crossentropy', optimizer=SGD(lr=0.01, momentum=0.9, nesterov=True))
 ```
 
-### Alternative implementation of MLP:
-
+You can now iterate on your training data in batches:
 ```python
-model = Sequential()
-model.add(Dense(20, 64, init='uniform', activation='tanh'))
-model.add(Dropout(0.5))
-model.add(Dense(64, 64, init='uniform', activation='tanh'))
-model.add(Dropout(0.5))
-model.add(Dense(64, 2, init='uniform', activation='softmax'))
-
-sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
-model.compile(loss='mean_squared_error', optimizer=sgd)
+model.fit(X_train, Y_train, nb_epoch=5, batch_size=32)
 ```
 
-### VGG-like convnet:
-
+Alternatively, you can feed batches to your model manually:
 ```python
-from keras.models import Sequential
-from keras.layers.core import Dense, Dropout, Activation, Flatten
-from keras.layers.convolutional import Convolution2D, MaxPooling2D
-from keras.optimizers import SGD
-
-model = Sequential()
-model.add(Convolution2D(32, 3, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-model.add(Dropout(0.25))
-
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(64, 64, 3, 3)) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-model.add(Dropout(0.25))
-
-model.add(Flatten())
-model.add(Dense(64*8*8, 256))
-model.add(Activation('relu'))
-model.add(Dropout(0.5))
-
-model.add(Dense(256, 10))
-model.add(Activation('softmax'))
-
-sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
-model.compile(loss='categorical_crossentropy', optimizer=sgd)
-
-model.fit(X_train, Y_train, batch_size=32, nb_epoch=1)
-
+model.train_on_batch(X_batch, Y_batch)
 ```
 
-### Sequence classification with LSTM:
-
+Evaluate your performance in one line:
 ```python
-from keras.models import Sequential
-from keras.layers.core import Dense, Dropout, Activation
-from keras.layers.embeddings import Embedding
-from keras.layers.recurrent import LSTM
-
-model = Sequential()
-model.add(Embedding(max_features, 256))
-model.add(LSTM(256, 128, activation='sigmoid', inner_activation='hard_sigmoid'))
-model.add(Dropout(0.5))
-model.add(Dense(128, 1))
-model.add(Activation('sigmoid'))
-
-model.compile(loss='binary_crossentropy', optimizer='rmsprop')
-
-model.fit(X_train, Y_train, batch_size=16, nb_epoch=10)
-score = model.evaluate(X_test, Y_test, batch_size=16)
+objective_score = model.evaluate(X_test, Y_test, batch_size=32)
 ```
 
-### Architecture for learning image captions with a convnet and a Gated Recurrent Unit:
-(word-level embedding, caption of maximum length 16 words).
-
-Note that getting this to actually "work" will require using a bigger convnet, initialized with pre-trained weights.
-Displaying readable results will also require an embedding decoder.
-
+Or generate predictions on new data:
 ```python
-max_caption_len = 16
-
-model = Sequential()
-model.add(Convolution2D(32, 3, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(64, 64, 3, 3)) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-
-model.add(Convolution2D(128, 64, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(128, 128, 3, 3)) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-
-model.add(Flatten())
-model.add(Dense(128*4*4, 256))
-model.add(Activation('relu'))
-model.add(Dropout(0.5))
-
-model.add(RepeatVector(max_caption_len)) 
-# the GRU below returns sequences of max_caption_len vectors of size 256 (our word embedding size)
-model.add(GRU(256, 256, return_sequences=True))
-
-model.compile(loss='mean_squared_error', optimizer='rmsprop')
-
-# "images" is a numpy array of shape (nb_samples, nb_channels=3, width, height) 
-# "captions" is a numpy array of shape (nb_samples, max_caption_len=16, embedding_dim=256)
-# captions are supposed already embedded (dense vectors).
-model.fit(images, captions, batch_size=16, nb_epoch=100)
-    
+classes = model.predict_classes(X_test, batch_size=32)
+proba = model.predict_proba(X_test, batch_size=32)
 ```
 
-In the examples folder, you will find example models for real datasets:
-- CIFAR10 small images classification: Convnet with realtime data augmentation
-- IMDB movie review sentiment classification: LSTM over sequences of words
-- Reuters newswires topic classification: Multilayer Perceptron (MLP)
-- MNIST handwritten digits classification: MLP & CNN
-- Character-level text generation with LSTM
+Building a network of LSTMs, a deep CNN, a Neural Turing Machine, a word2vec embedder or any other model is just as fast. The ideas behind deep learning are simple, so why should their implementation be painful?
 
-...and more.
+Have a look at these [starter examples](http://keras.io/examples/).
+
+In the [examples folder](https://github.com/fchollet/keras/tree/master/examples) of the repo, you will find more advanced models: question-answering with memory networks, text generation with stacked LSTMs, neural turing machines, etc.
 
 
-## Current capabilities
+------------------
 
-For complete coverage of the API, check out [the Keras documentation](http://keras.io).
-
-A few highlights: convnets, LSTM, GRU, word2vec-style embeddings, PReLU, batch normalization...
 
 ## Installation
 
@@ -191,12 +105,23 @@ Keras uses the following dependencies:
 
 - numpy, scipy
 - pyyaml
-- Theano
-    - See installation instructions: http://deeplearning.net/software/theano/install.html#install
 - HDF5 and h5py (optional, required if you use model saving/loading functions)
 - Optional but recommended if you use CNNs: cuDNN.
 
-Once you have the dependencies installed, cd to the Keras folder and run the install command:
+When using the Theano backend:
+- Theano
+    - [See installation instructions](http://deeplearning.net/software/theano/install.html#install).
+
+**Note**: You should use the latest version of Theano, not the PyPI version. Install it with:
+```
+sudo pip install git+git://github.com/Theano/Theano.git
+```
+
+When using the TensorFlow backend:
+- TensorFlow
+    - [See installation instructions](https://github.com/tensorflow/tensorflow#download-and-setup).
+
+To install, `cd` to the Keras folder and run the install command:
 ```
 sudo python setup.py install
 ```
@@ -206,11 +131,45 @@ You can also install Keras from PyPI:
 sudo pip install keras
 ```
 
+------------------
+
+
+## Switching from Theano to TensorFlow
+
+By default, Keras will use Theano as its tensor manipulation library. [Follow these instructions](http://keras.io/backend/) to configure the Keras backend.
+
+------------------
+
+
+## Support
+
+You can ask questions and join the development discussion on the [Keras Google group](https://groups.google.com/forum/#!forum/keras-users).
+
+------------------
+
+
+## Contribution Guidelines
+
+Keras welcomes all contributions from the community. 
+
+- Keep a pragmatic mindset and avoid bloat. Only add to the source if that is the only path forward.
+- New features should be documented. Make sure you update the documentation along with your Pull Request.
+- Any new function or class should have a proper docstring.
+- The documentation for every new feature should include a usage example in the form of a code snippet. 
+- All changes should be tested. Make sure any new feature you add has a corresponding unit test.
+- Please no Pull Requests about coding style.
+- Even if you don't contribute to the Keras source code, if you have an application of Keras that is concise and powerful, please consider adding it to our collection of [examples](https://github.com/fchollet/keras/tree/master/examples).
+
+
+------------------
+
+
 ## Why this name, Keras?
 
 Keras (κέρας) means _horn_ in Greek. It is a reference to a literary image from ancient Greek and Latin literature, first found in the _Odyssey_, where dream spirits (_Oneiroi_, singular _Oneiros_) are divided between those who deceive men with false visions, who arrive to Earth through a gate of ivory, and those who announce a future that will come to pass, who arrive through a gate of horn. It's a play on the words κέρας (horn) / κραίνω (fulfill), and ἐλέφας (ivory) / ἐλεφαίρομαι (deceive).
 
-Keras was developed as part of the research effort of project ONEIROS (Open-ended Neuro-Electronic Intelligent Robot Operating System).
+Keras was initially developed as part of the research effort of project ONEIROS (Open-ended Neuro-Electronic Intelligent Robot Operating System).
 
 >_"Oneiroi are beyond our unravelling --who can be sure what tale they tell? Not all that men look for comes to pass. Two gates there are that give passage to fleeting Oneiroi; one is made of horn, one of ivory. The Oneiroi that pass through sawn ivory are deceitful, bearing a message that will not be fulfilled; those that come out through polished horn have truth behind them, to be accomplished for men who see them."_ Homer, Odyssey 19. 562 ff (Shewring translation).
 
+------------------

docs/mkdocs.yml

@@ -14,6 +14,7 @@ pages:
 - Home: index.md
 - Index: documentation.md
 - Examples: examples.md
+- FAQ: faq.md
 - Optimizers: optimizers.md
 - Objectives: objectives.md
 - Models: models.md
@@ -23,6 +24,7 @@ pages:
 - Constraints: constraints.md
 - Callbacks: callbacks.md
 - Datasets: datasets.md
+- Visualization: visualization.md
 - Layers:
   - Core Layers: layers/core.md
   - Convolutional Layers: layers/convolutional.md

docs/sources/activations.md

@@ -6,12 +6,12 @@ Activations can either be used through an `Activation` layer, or through the `ac
 ```python
 from keras.layers.core import Activation, Dense
 
-model.add(Dense(64, 64, init='uniform'))
+model.add(Dense(64))
 model.add(Activation('tanh'))
 ```
 is equivalent to:
 ```python
-model.add(Dense(20, 64, init='uniform', activation='tanh'))
+model.add(Dense(64, activation='tanh'))
 ```
 
 You can also pass an element-wise Theano function as an activation:
@@ -20,7 +20,7 @@ You can also pass an element-wise Theano function as an activation:
 def tanh(x):
     return theano.tensor.tanh(x)
 
-model.add(Dense(20, 64, init='uniform', activation=tanh))
+model.add(Dense(64, activation=tanh))
 model.add(Activation(tanh))
 ```
 

docs/sources/backend.md

@@ -0,0 +1,74 @@
+# Keras backends
+
+## What is a "backend"?
+
+Keras is a model-level library, providing high-level building blocks for developing deep learning models. It does not handle itself low-level operations such as tensor products, convolutions and so on. Instead, it relies on a specialized, well-optimized tensor manipulation library to do so, serving as the "backend engine" of Keras. Rather than picking one single tensor library and making the implementation of Keras tied to that library, Keras handles the problem in a modular way, and several different backend engines can be plugged seamlessly into Keras.
+
+At this time, Keras has two backend implementations available: the **Theano** backend and the **TensorFlow** backend.
+
+- [Theano](http://deeplearning.net/software/theano/) is an open-source symbolic tensor manipulation framework developed by LISA/MILA Lab at Université de Montréal.
+- [TensorFlow](http://www.tensorflow.org/) is an open-source symbolic tensor manipulation framework developed by Google, Inc.
+
+## Switching from one backend to another
+
+If you have run Keras at least once, you will find the Keras configuration file at:
+
+`~/.keras/keras.json`
+
+If it isn't there, you can create it.
+
+It probably looks like this:
+
+`{"epsilon": 1e-07, "floatx": "float32", "backend": "theano"}`
+
+Simply change the field `backend` to either `"theano"` or `"tensorflow"`, and Keras will use the new configuration next time you run any Keras code.
+
+## Using the abstract Keras backend to write new code
+
+If you want the Keras modules you write to be compatible with both Theano and TensorFlow, you have to write them via the abstract Keras backend API. Here's an intro.
+
+You can import the backend module via:
+```python
+from keras import backend as K
+```
+
+This instantiates an input placeholder. It's equivalent to `tf.placeholder()` or `T.matrix()`, `T.tensor3()`, etc.
+
+```python
+input = K.placeholder(shape=(2, 4, 5))
+# also works:
+input = K.placeholder(shape=(None, 4, 5))
+# also works:
+input = K.placeholder(ndim=3)
+```
+
+This instantiates a shared variable. It's equivalent to `tf.variable()` or `theano.shared()`.
+
+```python
+val = np.random.random((3, 4, 5))
+var = K.variable(value=val)
+
+# all-zeros variable:
+var = K.ones(shape=(3, 4, 5))
+# all-ones:
+var = K.zeros(shape=(3, 4, 5))
+```
+
+Most tensor operations you will need can be done as you would in TensorFlow or Theano:
+
+```python
+a = b + c * K.abs(d)
+c = K.dot(a, K.transpose(b))
+a = K.sum(b, axis=2)
+a = K.softmax(b)
+a = concatenate([b, c], axis=-1)
+# etc...
+```
+
+For more information, see the code at `keras/backend/theano_backend.py` and `keras/backend/tensorflow_backend.py`.
+
+
+
+
+
+

docs/sources/callbacks.md

@@ -33,7 +33,10 @@ The `logs` dictionary will contain keys for quantities relevant to the current b
 keras.callbacks.ModelCheckpoint(filepath, verbose=0, save_best_only=False)
 ```
 
-Save the model after every epoch. If `save_best_only=True`, the latest best model according to the validation loss will not be overwritten. 
+Save the model after every epoch. If `save_best_only=True`, the latest best model according to the validation loss will not be overwritten.
+`filepath` can contain named formatting options, which will be filled the value of `epoch` and keys in `logs` (passed in `on_epoch_end`).
+
+For example: if `filepath` is `weights.{epoch:02d}-{val_loss:.2f}.hdf5`, then multiple files will be save with the epoch number and the validation loss.
 
 
 ```python
@@ -72,7 +75,7 @@ class LossHistory(keras.callbacks.Callback):
         self.losses.append(logs.get('loss'))
 
 model = Sequential()
-model.add(Dense(784, 10, init='uniform'))
+model.add(Dense(10, input_dim=784, init='uniform'))
 model.add(Activation('softmax'))
 model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
 
@@ -94,7 +97,7 @@ print history.losses
 from keras.callbacks import ModelCheckpoint
 
 model = Sequential()
-model.add(Dense(784, 10, init='uniform'))
+model.add(Dense(10, input_dim=784, init='uniform'))
 model.add(Activation('softmax'))
 model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
 

docs/sources/constraints.md

@@ -12,7 +12,7 @@ These layers expose 2 keyword arguments:
 
 ```python
 from keras.constraints import maxnorm
-model.add(Dense(64, 64, W_constraint = maxnorm(2)))
+model.add(Dense(64, W_constraint = maxnorm(2)))
 ```
 
 ## Available constraints

docs/sources/documentation.md

@@ -5,6 +5,8 @@
 - [Home](index.md)
 - [Index](documentation.md)
 - [Examples](examples.md)
+- [FAQ](faq.md)
+- [Backend](backend.md)
 
 ---
 

docs/sources/examples.md

@@ -1,7 +1,7 @@
 
 Here are a few examples to get you started!
 
-### Multilayer Perceptron (MLP)
+### Multilayer Perceptron (MLP):
 
 ```python
 from keras.models import Sequential
@@ -9,13 +9,16 @@ from keras.layers.core import Dense, Dropout, Activation
 from keras.optimizers import SGD
 
 model = Sequential()
-model.add(Dense(20, 64, init='uniform'))
+# Dense(64) is a fully-connected layer with 64 hidden units.
+# in the first layer, you must specify the expected input data shape:
+# here, 20-dimensional vectors.
+model.add(Dense(64, input_dim=20, init='uniform'))
 model.add(Activation('tanh'))
 model.add(Dropout(0.5))
-model.add(Dense(64, 64, init='uniform'))
+model.add(Dense(64, init='uniform'))
 model.add(Activation('tanh'))
 model.add(Dropout(0.5))
-model.add(Dense(64, 2, init='uniform'))
+model.add(Dense(2, init='uniform'))
 model.add(Activation('softmax'))
 
 sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
@@ -25,25 +28,21 @@ model.fit(X_train, y_train, nb_epoch=20, batch_size=16)
 score = model.evaluate(X_test, y_test, batch_size=16)
 ```
 
----
-
-### Alternative implementation of MLP
+### Alternative implementation of MLP:
 
 ```python
 model = Sequential()
-model.add(Dense(20, 64, init='uniform', activation='tanh'))
+model.add(Dense(64, input_dim=20, init='uniform', activation='tanh'))
 model.add(Dropout(0.5))
-model.add(Dense(64, 64, init='uniform', activation='tanh'))
+model.add(Dense(64, init='uniform', activation='tanh'))
 model.add(Dropout(0.5))
-model.add(Dense(64, 2, init='uniform', activation='softmax'))
+model.add(Dense(2, init='uniform', activation='softmax'))
 
 sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
 model.compile(loss='mean_squared_error', optimizer=sgd)
 ```
 
----
-
-### VGG-like convnet
+### VGG-like convnet:
 
 ```python
 from keras.models import Sequential
@@ -52,26 +51,29 @@ from keras.layers.convolutional import Convolution2D, MaxPooling2D
 from keras.optimizers import SGD
 
 model = Sequential()
-model.add(Convolution2D(32, 3, 3, 3, border_mode='full')) 
+# input: 100x100 images with 3 channels -> (3, 100, 100) tensors.
+# this applies 32 convolution filters of size 3x3 each.
+model.add(Convolution2D(32, 3, 3, border_mode='full', input_shape=(3, 100, 100)))
 model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
+model.add(Convolution2D(32, 3, 3))
 model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+model.add(MaxPooling2D(pool_size=(2, 2)))
 model.add(Dropout(0.25))
 
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full')) 
+model.add(Convolution2D(64, 3, 3, border_mode='valid'))
 model.add(Activation('relu'))
-model.add(Convolution2D(64, 64, 3, 3)) 
+model.add(Convolution2D(64, 3, 3))
 model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+model.add(MaxPooling2D(pool_size=(2, 2)))
 model.add(Dropout(0.25))
 
 model.add(Flatten())
-model.add(Dense(64*8*8, 256))
+# Note: Keras does automatic shape inference.
+model.add(Dense(256))
 model.add(Activation('relu'))
 model.add(Dropout(0.5))
 
-model.add(Dense(256, 10))
+model.add(Dense(10))
 model.add(Activation('softmax'))
 
 sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
@@ -81,9 +83,7 @@ model.fit(X_train, Y_train, batch_size=32, nb_epoch=1)
 
 ```
 
----
-
-### Sequence classification with LSTM
+### Sequence classification with LSTM:
 
 ```python
 from keras.models import Sequential
@@ -92,11 +92,10 @@ from keras.layers.embeddings import Embedding
 from keras.layers.recurrent import LSTM
 
 model = Sequential()
-# Add a mask_zero=True to the Embedding connstructor if 0 is a left-padding value in your data
-model.add(Embedding(max_features, 256))
-model.add(LSTM(256, 128, activation='sigmoid', inner_activation='hard_sigmoid'))
+model.add(Embedding(max_features, 256, input_length=maxlen))
+model.add(LSTM(output_dim=128, activation='sigmoid', inner_activation='hard_sigmoid'))
 model.add(Dropout(0.5))
-model.add(Dense(128, 1))
+model.add(Dense(1))
 model.add(Activation('sigmoid'))
 
 model.compile(loss='binary_crossentropy', optimizer='rmsprop')
@@ -105,59 +104,73 @@ model.fit(X_train, Y_train, batch_size=16, nb_epoch=10)
 score = model.evaluate(X_test, Y_test, batch_size=16)
 ```
 
----
+### Architecture for learning image captions with a convnet and a Gated Recurrent Unit:
+(word-level embedding, caption of maximum length 16 words).
 
-### Image captioning
-
-Architecture for learning image captions with a convnet and a Gated Recurrent Unit (word-level embedding, caption of maximum length 16 words).
-
-Note that getting this to actually "work" will require using a bigger convnet, initialized with pre-trained weights.
-Displaying readable results will also require an embedding decoder.
+Note that getting this to work well will require using a bigger convnet, initialized with pre-trained weights.
 
 ```python
 max_caption_len = 16
+vocab_size = 10000
 
-model = Sequential()
-model.add(Convolution2D(32, 3, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+# first, let's define an image model that
+# will encode pictures into 128-dimensional vectors.
+# it should be initialized with pre-trained weights.
+image_model = Sequential()
+image_model.add(Convolution2D(32, 3, 3, border_mode='full', input_shape=(3, 100, 100)))
+image_model.add(Activation('relu'))
+image_model.add(Convolution2D(32, 3, 3))
+image_model.add(Activation('relu'))
+image_model.add(MaxPooling2D(pool_size=(2, 2)))
 
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(64, 64, 3, 3)) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+image_model.add(Convolution2D(64, 3, 3, border_mode='full'))
+image_model.add(Activation('relu'))
+image_model.add(Convolution2D(64, 3, 3))
+image_model.add(Activation('relu'))
+image_model.add(MaxPooling2D(pool_size=(2, 2)))
 
-model.add(Convolution2D(128, 64, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(Convolution2D(128, 128, 3, 3)) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+image_model.add(Flatten())
+image_model.add(Dense(128))
 
-model.add(Flatten())
-model.add(Dense(128*4*4, 256))
-model.add(Activation('relu'))
-model.add(Dropout(0.5))
+# let's load the weights from a save file.
+image_model.load_weights('weight_file.h5')
 
-model.add(RepeatVector(max_caption_len)) 
-# the GRU below returns sequences of max_caption_len vectors of size 256 (our word embedding size)
-model.add(GRU(256, 256, return_sequences=True))
+# next, let's define a RNN model that encodes sequences of words
+# into sequences of 128-dimensional word vectors.
+language_model = Sequential()
+language_model.add(Embedding(vocab_size, 256, input_length=max_caption_len))
+language_model.add(GRU(output_dim=128, return_sequences=True))
+language_model.add(TimeDistributedDense(128))
 
-model.compile(loss='mean_squared_error', optimizer='rmsprop')
+# let's repeat the image vector to turn it into a sequence.
+image_model.add(RepeatVector(max_caption_len))
 
-# "images" is a numpy array of shape (nb_samples, nb_channels=3, width, height) 
-# "captions" is a numpy array of shape (nb_samples, max_caption_len=16, embedding_dim=256)
-# captions are supposed already embedded (dense vectors).
-model.fit(images, captions, batch_size=16, nb_epoch=100)
-    
+# the output of both models will be tensors of shape (samples, max_caption_len, 128).
+# let's concatenate these 2 vector sequences.
+model = Merge([image_model, language_model], mode='concat', concat_axis=-1)
+# let's encode this vector sequence into a single vector
+model.add(GRU(256, 256, return_sequences=False))
+# which will be used to compute a probability
+# distribution over what the next word in the caption should be!
+model.add(Dense(vocab_size))
+model.add(Activation('softmax'))
+
+model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
+
+# "images" is a numpy float array of shape (nb_samples, nb_channels=3, width, height).
+# "captions" is a numpy integer array of shape (nb_samples, max_caption_len)
+# containing word index sequences representing partial captions.
+# "next_words" is a numpy float array of shape (nb_samples, vocab_size)
+# containing a categorical encoding (0s and 1s) of the next word in the corresponding
+# partial caption.
+model.fit([images, partial_captions], next_words, batch_size=16, nb_epoch=100)
 ```
 
----
-
-In the [examples folder](https://github.com/fchollet/keras/tree/master/examples), you will find example models for real datasets:
-
-- CIFAR10 small images classification: Convnet with realtime data augmentation
+In the examples folder, you will find example models for real datasets:
+- CIFAR10 small images classification: Convolutional Neural Network (CNN) with realtime data augmentation
 - IMDB movie review sentiment classification: LSTM over sequences of words
-- Reuters newswires topic classification: Multilayer Perceptron
+- Reuters newswires topic classification: Multilayer Perceptron (MLP)
+- MNIST handwritten digits classification: MLP & CNN
+- Character-level text generation with LSTM
+
+...and more.

docs/sources/faq.md

@@ -0,0 +1,179 @@
+# Keras FAQ: Frequently Asked Keras Questions
+
+[How can I run Keras on GPU?](#how-can-i-run-keras-on-gpu)
+
+[How can I save a Keras model?](#how-can-i-save-a-keras-model)
+
+[Why is the training loss much higher than the testing loss?](#why-is-the-training-loss-much-higher-than-the-testing-loss)
+
+[How can I visualize the output of an intermediate layer?](#how-can-i-visualize-the-output-of-an-intermediate-layer)
+
+[Isn't there a bug with Merge or Graph related to input concatenation?](#isnt-there-a-bug-with-merge-or-graph-related-to-input-concatenation)
+
+[How can I use Keras with datasets that don't fit in memory?](#how-can-i-use-keras-with-datasets-that-dont-fit-in-memory)
+
+[How can I interrupt training when the validation loss isn't decreasing anymore?](#how-can-i-interrupt-training-when-the-validation-loss-isnt-decreasing-anymore)
+
+[How is the validation split computed?](#how-is-the-validation-split-computed)
+
+[Is the data shuffled during training?](#is-the-data-shuffled-during-training)
+
+[How can I record the training / validation loss / accuracy at each epoch?](#how-can-i-record-the-training-validation-loss-accuracy-at-each-epoch)
+
+---
+
+### How can I run Keras on GPU?
+
+Method 1: use Theano flags.
+```bash
+THEANO_FLAGS=device=gpu,floatX=float32 python my_keras_script.py
+```
+
+The name 'gpu' might have to be changed depending on your device's identifier (e.g. `gpu0`, `gpu1`, etc).
+
+Method 2: set up your `.theanorc`: [Instructions](http://deeplearning.net/software/theano/library/config.html)
+
+Method 3: manually set `theano.config.device`, `theano.config.floatX` at the beginning of your code:
+```python
+import theano
+theano.config.device = 'gpu'
+theano.config.floatX = 'float32'
+```
+
+---
+
+### How can I save a Keras model?
+
+*It is not recommended to use pickle or cPickle to save a Keras model.*
+
+If you only need to save the architecture of a model, and not its weights, you can do:
+
+```python
+# save as JSON
+json_string = model.to_json()
+
+# save as YAML
+yaml_string = model.to_yaml()
+```
+
+You can then build a fresh model from this data:
+
+```python
+# model reconstruction from JSON:
+from keras.models import model_from_json
+model = model_from_json(json_string)
+
+# model reconstruction from YAML
+model = model_from_yaml(yaml_string)
+```
+
+If you need to save the weights of a model, you can do so in HDF5:
+```python
+model.save_weights('my_model_weights.h5')
+```
+
+Assuming you have code for instantiating your model, you can then load the weights you saved into a model with the same architecture:
+
+```python
+model.load_weights('my_model_weights.h5')
+```
+
+This leads us to a way to save and reconstruct models from only serialized data:
+```python
+json_string = model.to_json()
+open('my_model_architecture.json', 'w').write(json_string)
+model.save_weights('my_model_weights.h5')
+
+# elsewhere...
+model = model_from_json(open('my_model_architecture.json').read())
+model.load_weights('my_model_weights.h5')
+```
+
+---
+
+### Why is the training loss much higher than the testing loss?
+
+A Keras model has two modes: training and testing. Regularization mechanisms, such as Dropout and L1/L2 weight regularization, are turned off at testing time.
+
+Besides, the training loss is the average of the losses over each batch of training data. Because your model is changing over time, the loss over the first batches of an epoch is generally higher than over the last batches. On the other hand, the testing loss for an epoch is computed using the model as it is at the end of the epoch, resulting in a lower loss.
+
+---
+
+### How can I visualize the output of an intermediate layer?
+
+You can build a Theano function that will return the output of a certain layer given a certain input, for example:
+
+```python
+# with a Sequential model
+get_3rd_layer_output = theano.function([model.layers[0].input], 
+                                       model.layers[3].get_output(train=False))
+layer_output = get_3rd_layer_output(X)
+
+# with a Graph model
+get_conv_layer_output = theano.function([model.inputs[i].input for i in model.input_order],
+                                        model.outputs['conv'].get_output(train=False),
+                                        on_unused_input='ignore')
+conv_output = get_conv_output(input_data_dict)
+```
+
+---
+
+### Isn't there a bug with Merge or Graph related to input concatenation?
+
+Yes, there was a known bug with tensor concatenation in Thenao that was fixed early 2015. 
+Please upgrade to the latest version of Theano:
+
+```bash
+sudo pip install git+git://github.com/Theano/Theano.git
+```
+
+---
+
+### How can I use Keras with datasets that don't fit in memory?
+
+You can do batch training using `model.train_on_batch(X, y)` and `model.test_on_batch(X, y)`. See the [models documentation](models.md).
+
+You can also see batch training in action in our [CIFAR10 example](https://github.com/fchollet/keras/blob/master/examples/cifar10_cnn.py).
+
+---
+
+### How can I interrupt training when the validation loss isn't decreasing anymore?
+
+You can use an `EarlyStopping` callback:
+
+```python
+from keras.callbacks import EarlyStopping
+early_stopping = EarlyStopping(monitor='val_loss', patience=2)
+model.fit(X, y, validation_split=0.2, callbacks=[early_stopping])
+```
+
+Find out more in the [callbacks documentation](callbacks.md).
+
+---
+
+### How is the validation split computed?
+
+If you set the `validation_split` arugment in `model.fit` to e.g. 0.1, then the validation data used will be the *last 10%* of the data. If you set it to 0.25, it will be the last 25% of the data, etc.
+
+
+---
+
+### Is the data shuffled during training?
+
+Yes, if the `shuffle` argument in `model.fit` is set to `True` (which is the default), the training data will be randomly shuffled at each epoch.
+
+Validation data isn't shuffled.
+
+---
+
+
+### How can I record the training / validation loss / accuracy at each epoch?
+
+The `model.fit` method returns an `History` callback, which has a `history` attribute containing the lists of successive losses / accuracies.
+
+```python
+hist = model.fit(X, y, validation_split=0.2)
+print(hist.history)
+```
+
+---

docs/sources/index.md

@@ -1,33 +1,38 @@
-# Keras: Theano-based Deep Learning library
+# Keras: Deep Learning library for Theano and TensorFlow
 
-## Overview
+## You have just found Keras.
 
-Keras is a minimalist, highly modular neural network library in the spirit of Torch, written in Python, that uses [Theano](http://deeplearning.net/software/theano/) under the hood for optimized tensor manipulation on GPU and CPU. It was developed with a focus on enabling fast experimentation. 
+Keras is a minimalist, highly modular neural networks library, written in Python and capable of running either on top of either [TensorFlow](https://github.com/tensorflow/tensorflow) or [Theano](https://github.com/Theano/Theano). It was developed with a focus on enabling fast experimentation. Being able to go from idea to result with the least possible delay is key to doing good research.
 
 Use Keras if you need a deep learning library that:
-
 - allows for easy and fast prototyping (through total modularity, minimalism, and extensibility).
 - supports both convolutional networks and recurrent networks, as well as combinations of the two.
 - supports arbitrary connectivity schemes (including multi-input and multi-output training).
 - runs seamlessly on CPU and GPU.
 
+Read the documentation at [Keras.io](http://keras.io).
+
+Keras is compatible with:
+    - __Python 2.7-3.5__ with the Theano backend
+    - __Python 2.7__ with the TensorFlow backend
+
+
+------------------
+
+
 ## Guiding principles
 
 - __Modularity.__ A model is understood as a sequence or a graph of standalone, fully-configurable modules that can be plugged together with as little restrictions as possible. In particular, neural layers, cost functions, optimizers, initialization schemes, activation functions, regularization schemes are all standalone modules that you can combine to create new models.
 
-- __Minimalism.__ Each module should be kept short and simple (<100 lines of code). Every piece of code should be transparent upon first reading. No black magic: it hurts iteration speed and ability to innovate.
+- __Minimalism.__ Each module should be kept short and simple. Every piece of code should be transparent upon first reading. No black magic: it hurts iteration speed and ability to innovate.
 
-- __Easy extensibility.__ New modules are dead simple to add (as new classes/functions), and existing modules provide ample examples. To be able to easily create new modules allows for total expressiveness, making Keras suitable for advanced research.
+- __Easy extensibility.__ New modules are dead simple to add (as new classes and functions), and existing modules provide ample examples. To be able to easily create new modules allows for total expressiveness, making Keras suitable for advanced research.
 
-- __Work with Python__. No separate models configuration files in a declarative format (like in Caffe or PyLearn2). Models are described in Python code, which is compact, easier to debug, and allows for ease of extensibility.
+- __Work with Python__. No separate models configuration files in a declarative format. Models are described in Python code, which is compact, easier to debug, and allows for ease of extensibility.
 
-## Code
 
-Find the code on Github: [fchollet/keras](https://github.com/fchollet/keras).
+------------------
 
-## License
-
-Keras is licensed under the [MIT license](http://opensource.org/licenses/MIT). 
 
 ## Getting started: 30 seconds to Keras
 
@@ -46,9 +51,9 @@ Stacking layers is as easy as `.add()`:
 ```python
 from keras.layers.core import Dense, Activation
 
-model.add(Dense(input_dim=100, output_dim=64, init="glorot_uniform"))
+model.add(Dense(output_dim=64, input_dim=100, init="glorot_uniform"))
 model.add(Activation("relu"))
-model.add(Dense(input_dim=64, output_dim=10, init="glorot_uniform"))
+model.add(Dense(output_dim=10, init="glorot_uniform"))
 model.add(Activation("softmax"))
 ```
 
@@ -57,7 +62,7 @@ Once your model looks good, configure its learning process with `.compile()`:
 model.compile(loss='categorical_crossentropy', optimizer='sgd')
 ```
 
-If you need to, you can further configure your optimizer. A core principle of Keras is make things things reasonably simple, while allowing the user to be fully in control when they need to (the ultimate control being the easy extensibility of the source code).
+If you need to, you can further configure your optimizer. A core principle of Keras is to make things reasonably simple, while allowing the user to be fully in control when they need to (the ultimate control being the easy extensibility of the source code).
 ```python
 from keras.optimizers import SGD
 model.compile(loss='categorical_crossentropy', optimizer=SGD(lr=0.01, momentum=0.9, nesterov=True))
@@ -86,55 +91,85 @@ proba = model.predict_proba(X_test, batch_size=32)
 
 Building a network of LSTMs, a deep CNN, a Neural Turing Machine, a word2vec embedder or any other model is just as fast. The ideas behind deep learning are simple, so why should their implementation be painful?
 
-Have a look at the [examples](examples.md).
+Have a look at these [starter examples](http://keras.io/examples/).
+
+In the [examples folder](https://github.com/fchollet/keras/tree/master/examples) of the repo, you will find more advanced models: question-answering with memory networks, text generation with stacked LSTMs, neural turing machines, etc.
+
+
+------------------
+
 
 ## Installation
 
 Keras uses the following dependencies:
 
-- __numpy__, __scipy__
-- __pyyaml__
-- __Theano__
-    - See [installation instructions](http://deeplearning.net/software/theano/install.html#install).
-- __HDF5__ and __h5py__ (optional, required if you use model saving/loading functions)
-- Optional but recommended if you use CNNs: __cuDNN__.
+- numpy, scipy
+- pyyaml
+- HDF5 and h5py (optional, required if you use model saving/loading functions)
+- Optional but recommended if you use CNNs: cuDNN.
 
-Once you have the dependencies installed, clone the repo:
-```bash
-git clone https://github.com/fchollet/keras.git
+When using the Theano backend:
+- Theano
+    - [See installation instructions](http://deeplearning.net/software/theano/install.html#install).
+
+**Note**: You should use the latest version of Theano, not the PyPI version. Install it with:
+```
+sudo pip install git+git://github.com/Theano/Theano.git
+```
+
+When using the TensorFlow backend:
+- TensorFlow
+    - [See installation instructions](https://github.com/tensorflow/tensorflow#download-and-setup).
+
+To install, `cd` to the Keras folder and run the install command:
 ```
-Go to the Keras folder and run the install command:
-```bash
-cd keras
 sudo python setup.py install
 ```
+
 You can also install Keras from PyPI:
 ```
 sudo pip install keras
 ```
 
+------------------
+
+
+## Switching from Theano to TensorFlow
+
+By default, Keras will use Theano as its tensor manipulation library. [Follow these instructions](http://keras.io/backend/) to configure the Keras backend.
+
+------------------
+
+
 ## Support
 
 You can ask questions and join the development discussion on the [Keras Google group](https://groups.google.com/forum/#!forum/keras-users).
 
+------------------
+
+
 ## Contribution Guidelines
 
 Keras welcomes all contributions from the community. 
 
 - Keep a pragmatic mindset and avoid bloat. Only add to the source if that is the only path forward.
 - New features should be documented. Make sure you update the documentation along with your Pull Request.
+- Any new function or class should have a proper docstring.
 - The documentation for every new feature should include a usage example in the form of a code snippet. 
 - All changes should be tested. Make sure any new feature you add has a corresponding unit test.
 - Please no Pull Requests about coding style.
 - Even if you don't contribute to the Keras source code, if you have an application of Keras that is concise and powerful, please consider adding it to our collection of [examples](https://github.com/fchollet/keras/tree/master/examples).
 
 
+------------------
+
+
 ## Why this name, Keras?
 
 Keras (κέρας) means _horn_ in Greek. It is a reference to a literary image from ancient Greek and Latin literature, first found in the _Odyssey_, where dream spirits (_Oneiroi_, singular _Oneiros_) are divided between those who deceive men with false visions, who arrive to Earth through a gate of ivory, and those who announce a future that will come to pass, who arrive through a gate of horn. It's a play on the words κέρας (horn) / κραίνω (fulfill), and ἐλέφας (ivory) / ἐλεφαίρομαι (deceive).
 
-Keras was developed as part of the research effort of project __ONEIROS__ (*Open-ended Neuro-Electronic Intelligent Robot Operating System*).
+Keras was initially developed as part of the research effort of project ONEIROS (Open-ended Neuro-Electronic Intelligent Robot Operating System).
 
-> _"Oneiroi are beyond our unravelling --who can be sure what tale they tell? Not all that men look for comes to pass. Two gates there are that give passage to fleeting Oneiroi; one is made of horn, one of ivory. The Oneiroi that pass through sawn ivory are deceitful, bearing a message that will not be fulfilled; those that come out through polished horn have truth behind them, to be accomplished for men who see them."_ 
+>_"Oneiroi are beyond our unravelling --who can be sure what tale they tell? Not all that men look for comes to pass. Two gates there are that give passage to fleeting Oneiroi; one is made of horn, one of ivory. The Oneiroi that pass through sawn ivory are deceitful, bearing a message that will not be fulfilled; those that come out through polished horn have truth behind them, to be accomplished for men who see them."_ Homer, Odyssey 19. 562 ff (Shewring translation).
 
-> -- Homer, Odyssey 19. 562 ff (Shewring translation).
+------------------

docs/sources/initializations.md

@@ -6,7 +6,7 @@ Initializations define the probability distribution used to set the initial rand
 The keyword arguments used for passing initializations to layers will depend on the layer. Usually it is simply `init`:
 
 ```python
-model.add(Dense(64, 64, init='uniform'))
+model.add(Dense(64, init='uniform'))
 ```
 
 ## Available initializations

docs/sources/layers/advanced_activations.md

@@ -7,7 +7,8 @@ keras.layers.advanced_activations.LeakyReLU(alpha=0.3)
 
 Special version of a Rectified Linear Unit that allows a small gradient when the unit is not active (`f(x) = alpha*x for x < 0`).
 
-- __Input shape__: This layer does not assume a specific input shape. As a result, it cannot be used as the first layer in a model.
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -19,38 +20,51 @@ Special version of a Rectified Linear Unit that allows a small gradient when the
 ## PReLU
 
 ```python
-keras.layers.advanced_activations.PReLU(input_shape)
+keras.layers.advanced_activations.PReLU()
 ```
 
 Parametrized linear unit. Similar to a LeakyReLU, where each input unit has its alpha coefficient, and where these coefficients are learned during training.
 
-- __Input shape__: Same as `input_shape`. This layer cannot be used as first layer in a model.
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
-- __Arguments__:
-    - __input_shape__: tuple.
-
 - __References__:
     - [Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification](http://arxiv.org/pdf/1502.01852v1.pdf)
 
 ---
 
+## ELU
+
+```python
+keras.layers.advanced_activations.ELU()
+```
+
+Exponential linear unit. Negative values pushes mean unit activations closer to zero, with the advantage of having a noise-robust deactivation state.
+
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
+
+- __Output shape__: Same as input.
+
+- __References__:
+    - [Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)](http://arxiv.org/pdf/1511.07289v1.pdf)
+
+---
+
 ## ParametricSoftplus
 
 ```python
-keras.layers.advanced_activations.ParametricSoftplus(input_shape)
+keras.layers.advanced_activations.ParametricSoftplus()
 ```
 
 Parametric Softplus of the form: (`f(x) = alpha * (1 + exp(beta * x))`). This is essentially a smooth version of ReLU where the parameters control the sharpness of the rectification. The parameters are initialized to more closely approximate a ReLU than the standard `softplus`: `alpha` initialized to `0.2` and `beta`  initialized to `5.0`. The parameters are fit separately for each hidden unit.
 
-- __Input shape__: Same as `input_shape`. This layer cannot be used as first layer in a model.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape=...` when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
-- __Arguments__:
-    - __input_shape__: tuple.
-
 - __References__:
     - [Inferring Nonlinear Neuronal Computation Based on Physiologically Plausible Inputs](http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1003143)
 
@@ -62,7 +76,8 @@ keras.layers.advanced_activations.ThresholdedLinear(theta)
 
 Parametrized linear unit. provides a threshold near zero where values are zeroed.
 
-- __Input shape__: Same as `input_shape`. This layer cannot be used as first layer in a model.
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -80,7 +95,7 @@ keras.layers.advanced_activations.ThresholdedReLu(theta)
 
 Parametrized rectified linear unit. provides a threshold near zero where values are zeroed.
 
-- __Input shape__: Same as `input_shape`. This layer cannot be used as first layer in a model.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape=...` when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 

docs/sources/layers/convolutional.md

@@ -2,69 +2,80 @@
 ## Convolution1D
 
 ```python
-keras.layers.convolutional.Convolution1D(input_dim, nb_filter, filter_length, 
-        init='uniform', activation='linear', weights=None, 
-        border_mode='valid', subsample_length=1, 
-        W_regularizer=None, b_regularizer=None, W_constraint=None, 
-        b_constraint=None)
+keras.layers.convolutional.Convolution1D(nb_filter, filter_length, 
+        init='uniform',
+        activation='linear',
+        weights=None, 
+        border_mode='valid',
+        subsample_length=1, 
+        W_regularizer=None, b_regularizer=None,
+        W_constraint=None, b_constraint=None,
+        input_dim=None, input_length=None)
 ```
 
-Convolution operator for filtering neighborhoods of one-dimensional inputs.
+Convolution operator for filtering neighborhoods of one-dimensional inputs. When using this layer as the first layer in a model, either provide the keyword argument `input_dim` (int, e.g. 128 for sequences of 128-dimensional vectors), or `input_shape` (tuple of integers, e.g. (10, 128) for sequences of 10 vectors of 128-dimensional vectors).
 
+- __Input shape__: 3D tensor with shape: `(samples, steps, input_dim)`.
 
-- __Input shape__: 3D tensor with shape: `(nb_samples, steps, input_dim)`.
-
-- __Output shape__: 3D tensor with shape: `(nb_samples, steps, nb_filter)`. `steps` value might have changed due to padding.
+- __Output shape__: 3D tensor with shape: `(samples, new_steps, nb_filter)`. `steps` value might have changed due to padding.
 
 - __Arguments__:
-    - __input_dim__: Number of channels/dimensions in the input.
     - __nb_filter__: Number of convolution kernels to use (dimensionality of the output).
     - __filter_length__: The extension (spatial or temporal) of each filter.
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
     - __activation__: name of activation function to use (see: [activations](../activations.md)), or alternatively, elementwise Theano function. If you don't specify anything, no activation is applied (ie. "linear" activation: a(x) = x).
     - __weights__: list of numpy arrays to set as initial weights.
-    - __border_mode__: 'valid' or 'full'. see scipy.signal.convolve2d.
+    - __border_mode__: 'valid' or 'same'.
     - __subsample_length__: factor by which to subsample output.
     - __W_regularizer__: instance of [WeightRegularizer](../regularizers.md) (eg. L1 or L2 regularization), applied to the main weights matrix.
     - __b_regularizer__: instance of [WeightRegularizer](../regularizers.md), applied to the bias.
     - __activity_regularizer__: instance of [ActivityRegularizer](../regularizers.md), applied to the network output.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the main weights matrix.
     - __b_constraint__: instance of the [constraints](../constraints.md) module, applied to the bias.
+    - __input_dim__: Number of channels/dimensions in the input. Either this argument or the keyword argument `input_shape` must be provided when using this layer as the first layer in a model.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
 
 ---
 
 ## Convolution2D
 
 ```python
-keras.layers.convolutional.Convolution2D(nb_filter, stack_size, nb_row, nb_col, 
-        init='glorot_uniform', activation='linear', weights=None, 
-        border_mode='valid', subsample=(1, 1),
-        W_regularizer=None, b_regularizer=None, W_constraint=None)
+keras.layers.convolutional.Convolution2D(nb_filter, nb_row, nb_col, 
+        init='glorot_uniform',
+        activation='linear',
+        weights=None, 
+        border_mode='valid',
+        subsample=(1, 1),
+        W_regularizer=None, b_regularizer=None,
+        W_constraint=None,
+        dim_ordering='th')
 ```
 
-Convolution operator for filtering windows of two-dimensional inputs. 
+Convolution operator for filtering windows of two-dimensional inputs. When using this layer as the first layer in a model, provide the keyword argument `input_shape` (tuple of integers, does not include the sample axis), e.g. `input_shape=(3, 128, 128)` for 128x128 RGB pictures.
 
-- __Input shape__: 4D tensor with shape: `(nb_samples, stack_size, nb_row, nb_col)`.
+- __Input shape__: 4D tensor with shape: `(samples, channels, rows, cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, rows, cols, channels)` if dim_ordering='tf'.
 
-- __Output shape__: 4D tensor with shape: `(nb_samples, nb_filter, nb_row, nb_col)`. `nb_row`, `nb_col` might have changed due to padding.
+- __Output shape__: 4D tensor with shape: `(samples, nb_filter, nb_row, nb_col)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, nb_row, nb_col, nb_filter)` if dim_ordering='tf'.
 
 
 - __Arguments__:
 
-    - __nb_filter__: Number of convolution kernels to use.
-    - __stack_size__: Number of channels in the input.
-    - __nb_row__: Number of rows in the convolution kernels
-    - __nb_col__: Number of columns in the convolution kernels
+    - __nb_filter__: Number of convolution filters to use.
+    - __nb_row__: Number of rows in the convolution kernel.
+    - __nb_col__: Number of columns in the convolution kernel.
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
     - __activation__: name of activation function to use (see: [activations](../activations.md)), or alternatively, elementwise Theano function. If you don't specify anything, no activation is applied (ie. "linear" activation: a(x) = x).
     - __weights__: list of numpy arrays to set as initial weights.
-    - __border_mode__: 'valid', 'full', or 'same'. [See scipy.signal.convolve2d](http://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.convolve2d.html).
+    - __border_mode__: 'valid' or 'same'.
     - __subsample__: tuple of length 2. Factor by which to subsample output. Also called strides elsewhere.
     - __W_regularizer__: instance of [WeightRegularizer](../regularizers.md) (eg. L1 or L2 regularization), applied to the main weights matrix.
     - __b_regularizer__: instance of [WeightRegularizer](../regularizers.md), applied to the bias.
     - __activity_regularizer__: instance of [ActivityRegularizer](../regularizers.md), applied to the network output.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the main weights matrix.
     - __b_constraint__: instance of the [constraints](../constraints.md) module, applied to the bias.
+    - __dim_ordering__: 'th' or 'tf'. In 'th' mode, the channels dimension (the depth) is at index 1, in 'tf' mode is it at index 3.
 
 
 ---
@@ -72,33 +83,120 @@ Convolution operator for filtering windows of two-dimensional inputs.
 ## MaxPooling1D
 
 ```python
-keras.layers.convolutional.MaxPooling1D(pool_length=2, stride=None, ignore_border=True)
+keras.layers.convolutional.MaxPooling1D(pool_length=2, stride=None, border_mode='valid')
 ```
 
-- __Input shape__: 3D tensor with shape: `(nb_samples, steps, dim)`.
+Max pooling operation for temporal data.
 
-- __Output shape__: 3D tensor with shape: `(nb_samples, downsampled_steps, dim)`.
+- __Input shape__: 3D tensor with shape: `(samples, steps, features)`.
+
+- __Output shape__: 3D tensor with shape: `(samples, downsampled_steps, features)`.
 
 - __Arguments__:
 
     - __pool_length__: factor by which to downscale. 2 will halve the input.
     - __stride__: integer or None. Stride value.
-    - __ignore_border__: boolean.
+    - __border_mode__: 'valid' or 'same'. **Note:** 'same' will only work with TensorFlow for the time being.
 
 ---
 
 ## MaxPooling2D
 
 ```python
-keras.layers.convolutional.MaxPooling2D(poolsize=(2, 2), ignore_border=True)
+keras.layers.convolutional.MaxPooling2D(pool_size=(2, 2), border_mode='valid', dim_ordering='th')
 ```
 
-- __Input shape__: 4D tensor with shape: `(nb_samples, stack_size, nb_row, nb_col)`.
+Max pooling operation for spatial data.
 
-- __Output shape__: 4D tensor with shape: `(nb_samples, stack_size, new_nb_row, new_nb_col)`.
+- __Input shape__: 4D tensor with shape: `(samples, channels, rows, cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, rows, cols, channels)` if dim_ordering='tf'.
+
+- __Output shape__: 4D tensor with shape: `(nb_samples, channels, pooled_rows, pooled_cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, pooled_rows, pooled_cols, channels)` if dim_ordering='tf'.
 
 - __Arguments__:
 
-    - __pool_size__: factor by which to downscale (vertical ds, horizontal ds). (2, 2) will halve the image in each dimension.
-    - __ignore_border__: boolean. When True, (5, 5) input with pool_size=(2, 2) will generate a (2, 2) output, (3, 3) otherwise.
+    - __pool_size__: tuple of 2 integers, factors by which to downscale (vertical, horizontal). (2, 2) will halve the image in each dimension.
+    - __strides__: tuple of 2 integers, or None. Strides values.
+    - __border_mode__: 'valid' or 'same'. **Note:** 'same' will only work with TensorFlow for the time being.
+    - __dim_ordering__: 'th' or 'tf'. In 'th' mode, the channels dimension (the depth) is at index 1, in 'tf' mode is it at index 3.
 
+
+---
+
+## UpSampling1D
+
+```python
+keras.layers.convolutional.UpSampling1D(length=2)
+```
+
+Repeats each temporal step `length` times along the time axis.
+
+- __Input shape__: 3D tensor with shape: `(samples, steps, features)`.
+
+- __Output shape__: 3D tensor with shape: `(samples, upsampled_steps, features)`.
+
+- __Arguments__:
+    - __length__: integer. Upsampling factor.
+
+---
+
+
+## UpSampling2D
+
+```python
+keras.layers.convolutional.UpSampling2D(size=(2, 2), dim_ordering='th')
+```
+
+Repeats the rows and columns of the data by size[0] and size[1] respectively.
+
+- __Input shape__: 4D tensor with shape: `(samples, channels, rows, cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, rows, cols, channels)` if dim_ordering='tf'.
+
+- __Output shape__: 4D tensor with shape: `(samples, channels, upsampled_rows, upsampled_cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, upsampled_rows, upsampled_cols, channels)` if dim_ordering='tf'.
+
+- __Arguments__:
+    - __size__: tuple of 2 integers. The upsampling factors for rows and columns.
+    - __dim_ordering__: 'th' or 'tf'. In 'th' mode, the channels dimension (the depth) is at index 1, in 'tf' mode is it at index 3.
+
+---
+
+
+## ZeroPadding1D
+
+```python
+keras.layers.convolutional.ZeroPaddding1D(padding=1)
+```
+
+Pads the input with zeros left and right along the time axis.
+
+- __Input shape__: 3D tensor with shape: `(nb_samples, steps, dim)`.
+
+- __Output shape__: 3D tensor with shape: `(nb_samples, padded_steps, dim)`.
+
+- __Arguments__:
+    - __padding__: integer, the size of the padding.
+
+---
+
+
+## ZeroPadding2D
+
+```python
+keras.layers.convolutional.ZeroPaddding2D(padding=(1, 1), dim_ordering='th')
+```
+
+Pads the rows and columns of the input with zeros, left and right.
+
+- __Input shape__: 4D tensor with shape: `(samples, channels, rows, cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, rows, cols, channels)` if dim_ordering='tf'.
+
+- __Output shape__: 4D tensor with shape: `(samples, channels, padded_rows, padded_cols)` if dim_ordering='th'
+or 4D tensor with shape: `(samples, padded_rows, padded_cols, channels)` if dim_ordering='tf'.
+
+- __Arguments__:
+    - __padding__: tuple of 2 integers, the size of the padding for rows and columns respectively.
+    - __dim_ordering__: 'th' or 'tf'. In 'th' mode, the channels dimension (the depth) is at index 1, in 'tf' mode is it at index 3.
+
+---

docs/sources/layers/core.md

@@ -76,8 +76,13 @@ get_config()
 
 ## Dense
 ```python
-keras.layers.core.Dense(input_dim, output_dim, init='glorot_uniform', activation='linear', weights=None \
-W_regularizer=None, b_regularizer=None, activity_regularizer=None, W_constraint=None, b_constraint=None)
+keras.layers.core.Dense(output_dim,
+                        init='glorot_uniform',
+                        activation='linear',
+                        weights=None,
+                        W_regularizer=None, b_regularizer=None, activity_regularizer=None,
+                        W_constraint=None, b_constraint=None,
+                        input_dim=None)
 ```
 
 Standard 1D fully-connect layer. 
@@ -88,7 +93,6 @@ Standard 1D fully-connect layer.
 
 - __Arguments__:
 
-    - __input_dim__: int >= 0. 
     - __output_dim__: int >= 0. 
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
     - __activation__: name of activation function to use (see: [activations](../activations.md)), or alternatively, elementwise Theano function. If you don't specify anything, no activation is applied (ie. "linear" activation: a(x) = x).
@@ -98,21 +102,26 @@ Standard 1D fully-connect layer.
     - __activity_regularizer__: instance of [ActivityRegularizer](../regularizers.md), applied to the network output.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the main weights matrix.
     - __b_constraint__: instance of the [constraints](../constraints.md) module, applied to the bias.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model. 
 
 ---
 
 ## TimeDistributedDense
 ```python
-keras.layers.core.TimeDistributedDense(input_dim, output_dim, init='glorot_uniform', activation='linear', weights=None \
-W_regularizer=None, b_regularizer=None, activity_regularizer=None, W_constraint=None, b_constraint=None)
+keras.layers.core.TimeDistributedDense(output_dim,
+                                       init='glorot_uniform',
+                                       activation='linear',
+                                       weights=None
+                                       W_regularizer=None, b_regularizer=None, activity_regularizer=None,
+                                       W_constraint=None, b_constraint=None,
+                                       input_dim=None, input_length=None)
 ```
 
 Fully-connected layer distributed over the time dimension. Useful after a recurrent network set to `return_sequences=True`.
 
-- __Input shape__: 3D tensor with shape: `(nb_samples, nb_timesteps, input_dim)`.
+- __Input shape__: 3D tensor with shape: `(nb_samples, timesteps, input_dim)`.
 
 - __Arguments__:
-    - __input_dim__: int >= 0. 
     - __output_dim__: int >= 0. 
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
     - __activation__: name of activation function to use (see: [activations](../activations.md)), or alternatively, elementwise Theano function. If you don't specify anything, no activation is applied (ie. "linear" activation: a(x) = x).
@@ -122,12 +131,14 @@ Fully-connected layer distributed over the time dimension. Useful after a recurr
     - __activity_regularizer__: instance of [ActivityRegularizer](../regularizers.md), applied to the network output.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the main weights matrix.
     - __b_constraint__: instance of the [constraints](../constraints.md) module, applied to the bias.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
 
 - __Example__:
 ```python
-# input shape: (nb_samples, nb_timesteps, 10)
-model.add(LSTM(10, 5, return_sequences=True)) # output shape: (nb_samples, nb_timesteps, 5)
-model.add(TimeDistributedDense(5, 10)) # output shape: (nb_samples, nb_timesteps, 10)
+# input shape: (nb_samples, timesteps, 10)
+model.add(LSTM(5, return_sequences=True, input_dim=10)) # output shape: (nb_samples, timesteps, 5)
+model.add(TimeDistributedDense(15)) # output shape: (nb_samples, timesteps, 15)
 ```
 
 
@@ -151,7 +162,7 @@ A customizable autoencoder model. If `output_reconstruction = True` then dim(inp
 
     - __decoder__: A [layer](./) or [layer container](./containers.md).
     
-    - __output_reconstruction__: If this is False the when .predict() is called the output is the deepest hidden layer's activation. Otherwise the output of the final decoder layer is presented. Be sure your validation data confirms to this logic if you decide to use any.
+    - __output_reconstruction__: If this is False, then when .predict() is called, the output is the deepest hidden layer's activation. Otherwise, the output of the final decoder layer is presented. Be sure your validation data conforms to this logic if you decide to use any.
     
     - __weights__: list of numpy arrays to set as initial weights. The list should have 1 element, of shape `(input_dim, output_dim)`.
 
@@ -160,8 +171,8 @@ A customizable autoencoder model. If `output_reconstruction = True` then dim(inp
 from keras.layers import containers
 
 # input shape: (nb_samples, 32)
-encoder = containers.Sequential([Dense(32, 16), Dense(16, 8)])
-decoder = containers.Sequential([Dense(8, 16), Dense(16, 32)])
+encoder = containers.Sequential([Dense(16, input_dim=32), Dense(8)])
+decoder = containers.Sequential([Dense(16, input_dim=8), Dense(32)])
 
 autoencoder = Sequential()
 autoencoder.add(AutoEncoder(encoder=encoder, decoder=decoder, output_reconstruction=False))
@@ -176,7 +187,8 @@ keras.layers.core.Activation(activation)
 ```
 Apply an activation function to the input. 
 
-- __Input shape__: This layer does not assume a specific input shape. As a result, it cannot be used as the first layer in a model.
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -193,7 +205,8 @@ keras.layers.core.Dropout(p)
 ```
 Apply dropout to the input. Dropout consists in randomly setting a fraction `p` of input units to 0 at each update during training time, which helps prevent overfitting. Reference: [Dropout: A Simple Way to Prevent Neural Networks from Overfitting](http://www.cs.toronto.edu/~rsalakhu/papers/srivastava14a.pdf)
 
-- __Input shape__: This layer does not assume a specific input shape. 
+
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -206,24 +219,25 @@ Apply dropout to the input. Dropout consists in randomly setting a fraction `p`
 
 ## Reshape
 ```python
-keras.layers.core.Reshape(*dims)
+keras.layers.core.Reshape(dims)
 ```
 
 Reshape the input to a new shape containing the same number of units. 
 
-- __Input shape__: This layer does not assume a specific input shape. 
 
-- __Output shape__: `(nb_samples, *dims)`.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
+
+- __Output shape__: `(nb_samples, dims)`.
 
 - __Arguments__:
 
-    - *dims: integers. Dimensions of the new shape.
+    - dims: tuple of integers. Dimensions of the new shape.
 
 - __Example__:
 ```python
 # input shape: (nb_samples, 10)
-model.add(Dense(10, 100)) # output shape: (nb_samples, 100)
-model.add(Reshape(10, 10))  # output shape: (nb_samples, 10, 10)
+model.add(Dense(100, input_dim=10)) # output shape: (nb_samples, 100)
+model.add(Reshape(dims=(10, 10)))  # output shape: (nb_samples, 10, 10)
 ```
 
 ---
@@ -235,7 +249,7 @@ keras.layers.core.Flatten()
 
 Convert a nD input to 1D. 
 
-- __Input shape__: (nb_samples, *). This layer cannot be used as the first layer in a model.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: `(nb_samples, nb_input_units)`.
 
@@ -250,7 +264,7 @@ Repeat the 1D input n times. Dimensions of input are assumed to be `(nb_samples,
 
 Note that the output is still a single tensor; `RepeatVector` does not split the data flow.
 
-- __Input shape__: This layer does not assume a specific input shape. This layer cannot be used as the first layer in a model.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: `(nb_samples, n, input_dims)`.
 
@@ -265,18 +279,18 @@ keras.layers.core.Permute(dims)
 ```
 Permute the dimensions of the input data according to the given tuple. Sometimes useful for connecting RNNs and convnets together.
 
-- __Input shape: This layer does not assume a specific input shape.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
-- __Output shape: Same as the input shape, but with the dimensions re-ordered according to the ordering specified by the tuple.
+- __Output shape__: Same as the input shape, but with the dimensions re-ordered according to the ordering specified by the tuple.
 
-- __Argument: tuple specifying the permutation scheme (e.g. `(2, 1)` permutes the first and second dimension of the input).
+- __Argument__: tuple specifying the permutation scheme (e.g. `(2, 1)` permutes the first and second dimension of the input).
 
 - __Example__:
 ```python
 # input shape: (nb_samples, 10)
-model.add(Dense(10, 50)) # output shape: (nb_samples, 50)
-model.add(Reshape(10, 5)) # output shape: (nb_samples, 10, 5)
-model.add(Permute((2, 1))) #output shape: (nb_samples, 5, 10)
+model.add(Dense(50, input_dim=10)) # output shape: (nb_samples, 50)
+model.add(Reshape(dims=(10, 5))) # output shape: (nb_samples, 10, 5)
+model.add(Permute(dims=(2, 1))) #output shape: (nb_samples, 5, 10)
 ```
 
 ---
@@ -294,8 +308,12 @@ This layer can be used, for instance, to induce activation sparsity in the previ
 
 ## MaxoutDense
 ```python
-keras.layers.core.MaxoutDense(input_dim, output_dim, nb_feature=4, init='glorot_uniform', weights=None, \
-        W_regularizer=None, b_regularizer=None, activity_regularizer=None, W_constraint=None, b_constraint=None)
+keras.layers.core.MaxoutDense(output_dim, nb_feature=4,
+        init='glorot_uniform',
+        weights=None,
+        W_regularizer=None, b_regularizer=None, activity_regularizer=None,
+        W_constraint=None, b_constraint=None,
+        input_dim=None)
 ```
 
 A dense maxout layer. A `MaxoutDense` layer takes the element-wise maximum of `nb_feature` `Dense(input_dim, output_dim)` linear layers. This allows the layer to learn a convex, piecewise linear activation function over the inputs. See [this paper](http://arxiv.org/pdf/1302.4389.pdf) for more details. Note that this is a *linear* layer -- if you wish to apply activation function (you shouldn't need to -- they are universal function approximators), an `Activation` layer must be added after.
@@ -306,7 +324,6 @@ A dense maxout layer. A `MaxoutDense` layer takes the element-wise maximum of `n
 
 - __Arguments__:
 
-    - __input_dim__: int >= 0. 
     - __output_dim__: int >= 0. 
     - __nb_feature__: int >= 0. the number of features to create for the maxout. This is equivalent to the number of piecewise elements to be allowed for the activation function. 
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
@@ -316,40 +333,46 @@ A dense maxout layer. A `MaxoutDense` layer takes the element-wise maximum of `n
     - __activity_regularizer__: instance of [ActivityRegularizer](../regularizers.md), applied to the network output.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the main weights matrix.
     - __b_constraint__: instance of the [constraints](../constraints.md) module, applied to the bias.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model.
 
 ```python
 # input shape: (nb_samples, 10)
-model.add(Dense(10, 100)) # output shape: (nb_samples, 100)
-model.add(MaxoutDense(100, 100, nb_feature=10)) # output shape: (nb_samples, 100)
-model.add(RepeatVector(2))  # output shape: (nb_samples, 2, 10)
+model.add(Dense(100, input_dim=10)) # output shape: (nb_samples, 100)
+model.add(MaxoutDense(50, nb_feature=10)) # output shape: (nb_samples, 50)
 ```
 
 ## Merge
 ```python
-keras.layers.core.Merge(models, mode='sum')
+keras.layers.core.Merge(layers, mode='sum', concat_axis=-1, dot_axes=-1)
 ```
 
-Merge the output of a list of layers (or containers) into a single tensor, following one of two modes: `sum` or `concat`. 
+Merge the output of a list of layers (or containers) into a single tensor.
 
 - __Arguments__:
     - __layers__: List of layers or [containers](/layers/containers/).
-    - __mode__: String, one of `{'sum', 'concat'}`. `sum` will simply sum the outputs of the layers (therefore all layers should have an output with the same shape). `concat` will concatenate the outputs along the last dimension (therefore all layers should have an output that only differ along the last dimension). 
+    - __mode__: String, one of `{'sum', 'mul', 'concat', 'ave', 'dot'}`. `sum`, `mul` and `ave` will simply sum/multiply/average the outputs of the layers (therefore all layers should have an output with the same shape). `concat` will concatenate the outputs along the dimension specified by `concate_axis` (therefore all layers should have an output that only differ along this dimension). `dot` will dot tensor contraction on the axes specified by `dot_axes` (see [the Numpy documentation](http://docs.scipy.org/doc/numpy-1.10.1/reference/generated/numpy.tensordot.html) for more details).
+    - __concat_axis__: axis to use in `concat` mode.
+    - __dot_axes__: axis or axes to use in `dot` mode (see [the Numpy documentation](http://docs.scipy.org/doc/numpy-1.10.1/reference/generated/numpy.tensordot.html) for more details).
+
+
+- __Notes__:
+    - `dot` mode only works with Theano for the time being.
 
 - __Example__:
 
 ```python
 left = Sequential()
-left.add(Dense(784, 50))
+left.add(Dense(50, input_shape=(784,)))
 left.add(Activation('relu'))
 
 right = Sequential()
-right.add(Dense(784, 50))
+right.add(Dense(50, input_shape=(784,)))
 right.add(Activation('relu'))
 
 model = Sequential()
 model.add(Merge([left, right], mode='sum'))
 
-model.add(Dense(50, 10))
+model.add(Dense(10))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
@@ -368,3 +391,87 @@ Given an input of dimensions `(nb_samples, timesteps, input_dim)`, return the in
 - __Input shape__: 3D tensor with shape: `(nb_samples, timesteps, features)`.
 
 - __Output shape__: 3D tensor with shape: `(nb_samples, timesteps, features)`.
+
+- __Notes__: Masking only works in Theano for the time being.
+
+## Lambda
+```python
+keras.layers.core.Lambda(function, output_shape=None)
+```
+
+Used for evaluating an arbitrary Theano expression on the output of the previous layer.
+
+- __Input shape__: Arbitrary. Use the keyword argument input_shape (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
+
+- __Output shape__: Specified by the `output_shape` argument.
+
+- __Arguments__:
+
+    - __function__: The expression to be evaluated. Takes one argument: the output of the previous layer.
+    - __output_shape__: Shape of the tensor returned by `function`. Should be a shape tuple (not including the samples dimension) or a function of the full input shape tuple (including samples dimension).
+
+- __Example__:
+
+```python
+# custom softmax function
+def sharp_softmax(X, beta=1.5):
+    return theano.tensor.nnet.softmax(X * beta)
+
+def output_shape(input_shape):
+    # here input_shape includes the samples dimension
+    return input_shape  # shape is unchanged
+
+model = Sequential()
+model.add(Dense(input_dim=10, output_dim=10))
+model.add(Lambda(sharp_softmax, output_shape))
+model.add(Dense(1))
+model.add(Activation('sigmoid'))
+```
+
+
+## LambdaMerge
+```python
+keras.layers.core.LambdaMerge(layers, function, output_shape=None)
+```
+
+Merge the output of a list of layers (or containers) into a single tensor, using an arbitrary Theano expression.
+
+- __Arguments__:
+    - __layers__: List of layers or [containers](/layers/containers/).
+    - __function__: The expression to be evaluated. Takes one argument: the list of input tensors.
+    - __output_shape__: Shape of the tensor returned by `function`. Should be a shape tuple (not including samples dimension) or a function of the list of input shape tuples (including samples dimension).
+
+- __Example__:
+
+```python
+# root mean square function
+def rms(inputs):
+    # inputs is a list of tensors
+    s = inputs[0] ** 2
+    for i in range(1, len(inputs)):
+        s += inputs[i] ** 2
+    s /= len(inputs)
+    s = theano.tensor.sqrt(s)
+    # return a single tensor
+    return s
+
+def output_shape(input_shapes):
+    # return the shape of the first tensor
+    return input_shapes[0]
+
+left = Sequential()
+left.add(Dense(input_dim=10, output_dim=10))
+left.add(Activation('sigmoid'))
+
+right = Sequential()
+right.add(Dense(input_dim=10, output_dim=10))
+right.add(Activation('sigmoid'))
+
+model = Sequential()
+model.add(LambdaMerge([left, right], rms, output_shape))
+
+model.add(Dense(1))
+model.add(Activation('sigmoid'))
+```
+
+---

docs/sources/layers/embeddings.md

@@ -2,48 +2,30 @@
 ## Embedding
 
 ```python
-keras.layers.embeddings.Embedding(input_dim, output_dim, init='uniform', weights=None, W_regularizer=None, W_constraint=None, mask_zero=False)
+keras.layers.embeddings.Embedding(input_dim, output_dim,
+                                  init='uniform',
+                                  weights=None,
+                                  W_regularizer=None, W_constraint=None,
+                                  mask_zero=False,
+                                  input_length=None)
 ```
 
 Turn positive integers (indexes) into denses vectors of fixed size,
 eg. `[[4], [20]] -> [[0.25, 0.1], [0.6, -0.2]]`
 
-- __Input shape__: 2D tensor with shape: `(nb_samples, maxlen)`.
+- __Input shape__: 2D tensor with shape: `(nb_samples, sequence_length)`.
 
-- __Output shape__: 3D tensor with shape: `(nb_samples, maxlen, output_dim)`.
+- __Output shape__: 3D tensor with shape: `(nb_samples, sequence_length, output_dim)`.
 
 - __Arguments__:
 
-    - __input_dim__: int >= 0. Size of the vocabulary, ie. 1+maximum integer index occuring in the input data.
+    - __input_dim__: int >= 0. Size of the vocabulary, ie. 1+maximum integer index occurring in the input data.
     - __output_dim__: int >= 0. Dimension of the dense embedding.
     - __init__: name of initialization function for the weights of the layer (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
     - __weights__: list of numpy arrays to set as initial weights. The list should have 1 element, of shape `(input_dim, output_dim)`.
     - __W_regularizer__: instance of the [regularizers](../regularizers.md) module (eg. L1 or L2 regularization), applied to the embedding matrix.
     - __W_constraint__: instance of the [constraints](../constraints.md) module (eg. maxnorm, nonneg), applied to the embedding matrix.
 	- __mask_zero__: Whether or not the input value 0 is a special "padding" value that should be masked out. This is useful for [recurrent layers](recurrent.md) which may take variable length input. If this is `True` then all subsequent layers in the model need to support masking or an exception will be raised.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
 
-
-## WordContextProduct
-
-```python
-keras.layers.embeddings.WordContextProduct(input_dim, proj_dim=128, 
-        init='uniform', activation='sigmoid', weights=None)
-```
-
-This layer turns a pair of words (a pivot word + a context word, ie. a word from the same context as a pivot, or a random, out-of-context word), indentified by their indices in a vocabulary, into two dense reprensentations (word representation and context representation).
-
-Then it returns `activation(dot(pivot_embedding, context_embedding))`, which can be trained to encode the probability of finding the context word in the context of the pivot word (or reciprocally depending on your training procedure).
-
-For more context, see Mikolov et al.: [Efficient Estimation of Word reprensentations in Vector Space](http://arxiv.org/pdf/1301.3781v3.pdf)
-
-- __Input shape__: 2D tensor with shape: `(nb_samples, 2)`.
-
-- __Output shape__: 2D tensor with shape: `(nb_samples, 1)`.
-
-- __Arguments__:
-
-    - __input_dim__: int >= 0. Size of the vocabulary, ie. 1+maximum integer index occuring in the input data.
-    - __proj_dim__: int >= 0. Dimension of the dense embedding used internally.
-    - __init__: name of initialization function for the embeddings (see: [initializations](../initializations.md)), or alternatively, Theano function to use for weights initialization. This parameter is only relevant if you don't pass a `weights` argument.
-    - __activation__: name of activation function to use (see: [activations](../activations.md)), or alternatively, elementwise Theano function.
-    - __weights__: list of numpy arrays to set as initial weights. The list should have 2 element, both of shape `(input_dim, proj_dim)`. The first element is the word embedding weights, the second one is the context embedding weights.
+---

docs/sources/layers/noise.md

@@ -6,9 +6,9 @@ keras.layers.noise.GaussianNoise(sigma)
 ```
 Apply to the input an additive zero-centred gaussian noise with standard deviation `sigma`. This is useful to mitigate overfitting (you could see it as a kind of random data augmentation). Gaussian Noise (GS) is a natural choice as corruption process for real valued inputs.
 
-The Gaussian noise is only added at training time.
+Only active at training time.
 
-- __Input shape__: This layer does not assume a specific input shape. 
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -24,11 +24,9 @@ keras.layers.noise.GaussianDropout(p)
 ```
 Apply to the input an multiplicative one-centred gaussian noise with standard deviation `sqrt(p/(1-p))`. p refers to drop probability to match Dropout layer syntax. 
 
-http://www.cs.toronto.edu/~rsalakhu/papers/srivastava14a.pdf
+Only active at training time.
 
-The Gaussian noise is only used at training time.
-
-- __Input shape__: This layer does not assume a specific input shape. 
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
@@ -36,3 +34,4 @@ The Gaussian noise is only used at training time.
 
     - __p__: float, drop probability as with Dropout.
 
+

docs/sources/layers/normalization.md

@@ -2,17 +2,16 @@
 ## BatchNormalization
 
 ```python
-keras.layers.normalization.BatchNormalization(input_shape, epsilon=1e-6, weights=None)
+keras.layers.normalization.BatchNormalization(epsilon=1e-6, weights=None)
 ```
 
 Normalize the activations of the previous layer at each batch.
 
-- __Input shape__: Same as `input_shape`. This layer cannot be used as first layer in a model.
+- __Input shape__: Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model.
 
 - __Output shape__: Same as input.
 
-- __Arguments__:
-    - __input_shape__: tuple.
+- __Arguments__: 
     - __epsilon__: small float > 0. Fuzz parameter.
     - __weights__: Initialization weights. List of 2 numpy arrays, with shapes: `[(input_shape,), (input_shape,)]`
 

docs/sources/layers/recurrent.md

@@ -2,9 +2,14 @@
 ## SimpleRNN
 
 ```python
-keras.layers.recurrent.SimpleRNN(input_dim, output_dim, 
-        init='glorot_uniform', inner_init='orthogonal', activation='sigmoid', weights=None,
-        truncate_gradient=-1, return_sequences=False)
+keras.layers.recurrent.SimpleRNN(output_dim,
+        init='glorot_uniform', inner_init='orthogonal',
+        activation='sigmoid',
+        weights=None,
+        return_sequences=False,
+        go_backwards=False,
+        stateful=False,
+        input_dim=None, input_length=None)
 ```
 Fully connected RNN where output is to fed back to input. 
 
@@ -14,56 +19,21 @@ Fully connected RNN where output is to fed back to input.
     - if `return_sequences`: 3D tensor with shape: `(nb_samples, timesteps, output_dim)`.
     - else: 2D tensor with shape: `(nb_samples, output_dim)`.
 
-- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to `True`.
+- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to `True`. **Note:** for the time being, masking in only supported with Theano.
+
+- __Notes__: When using the TensorFlow backend, the number of timesteps used must be fixed. Make sure to pass an `input_length` int argument or a complete `input_shape` tuple argument.
 
 
 - __Arguments__:
-    - __input_dim__: dimension of the input.
     - __output_dim__: dimension of the internal projections and the final output.
     - __init__: weight initialization function. Can be the name of an existing function (str), or a Theano function (see: [initializations](../initializations.md)).
     - __activation__: activation function. Can be the name of an existing function (str), or a Theano function (see: [activations](../activations.md)).
     - __weights__: list of numpy arrays to set as initial weights. The list should have 3 elements, of shapes: `[(input_dim, output_dim), (output_dim, output_dim), (output_dim,)]`.
-    - __truncate_gradient__: Number of steps to use in truncated BPTT. See: [Theano "scan"](http://deeplearning.net/software/theano/library/scan.html).
-    - __return_sequences__: Boolean. Whether to return the last output in the output sequence, or the full sequence.
-
----
-
-## SimpleDeepRNN
-
-```python
-keras.layers.recurrent.SimpleDeepRNN(input_dim, output_dim, depth=3,
-        init='glorot_uniform', inner_init='orthogonal', 
-        activation='sigmoid', inner_activation='hard_sigmoid',
-        weights=None, truncate_gradient=-1, return_sequences=False)
-```
-Fully connected RNN where the output of multiple timesteps (up to "depth" steps in the past) is fed back to the input: 
-
-```
-output = activation( W.x_t + b + inner_activation(U_1.h_tm1) + inner_activation(U_2.h_tm2) + ... )
-```
-
-Not a particularly useful model, included for demonstration purposes.
-
-- __Input shape__: 3D tensor with shape: `(nb_samples, timesteps, input_dim)`.
-
-- __Output shape__:
-    - if `return_sequences`: 3D tensor with shape: `(nb_samples, timesteps, output_dim)`.
-    - else: 2D tensor with shape: `(nb_samples, output_dim)`.
-
-- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to `True`.
-
-
-- __Arguments__:
-    - __input_dim__: dimension of the input.
-    - __output_dim__: dimension of the internal projections and the final output.
-    - __depth__: int >= 1. Lookback depth (eg. depth=1 is equivalent to SimpleRNN).
-    - __init__: weight initialization function for the output cell. Can be the name of an existing function (str), or a Theano function (see: [initializations](../initializations.md)).
-    - __inner_init__: weight initialization function for the inner cells.
-    - __activation__: activation function for the output. Can be the name of an existing function (str), or a Theano function (see: [activations](../activations.md)).
-    - __inner_activation__: activation function for the inner cells.
-    - __weights__: list of numpy arrays to set as initial weights. The list should have depth+2 elements.
-    - __truncate_gradient__: Number of steps to use in truncated BPTT. See: [Theano "scan"](http://deeplearning.net/software/theano/library/scan.html).
     - __return_sequences__: Boolean. Whether to return the last output in the output sequence, or the full sequence.
+    - __go_backwards__: Boolean (default False). If True, rocess the input sequence backwards.
+    - __stateful__: Boolean (default False). If True, the last state for each sample at index i in a batch will be used as initial state for the sample of index i in the following batch.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
 
 
 ---
@@ -71,10 +41,13 @@ Not a particularly useful model, included for demonstration purposes.
 ## GRU
 
 ```python
-keras.layers.recurrent.GRU(input_dim, output_dim=128, 
+keras.layers.recurrent.GRU(output_dim,
         init='glorot_uniform', inner_init='orthogonal',
         activation='sigmoid', inner_activation='hard_sigmoid',
-        weights=None, truncate_gradient=-1, return_sequences=False)
+        return_sequences=False,
+        go_backwards=False,
+        stateful=False,
+        input_dim=None, input_length=None)
 ```
 
 Gated Recurrent Unit - Cho et al. 2014.
@@ -85,18 +58,23 @@ Gated Recurrent Unit - Cho et al. 2014.
     - if `return_sequences`: 3D tensor with shape: `(nb_samples, timesteps, output_dim)`.
     - else: 2D tensor with shape: `(nb_samples, output_dim)`.
 
-- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to true.
+- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to true. **Note:** for the time being, masking in only supported with Theano.
+
+- __Notes__: When using the TensorFlow backend, the number of timesteps used must be fixed. Make sure to pass an `input_length` int argument or a complete `input_shape` tuple argument.
 
 - __Arguments__:
-    - __input_dim__: dimension of the input.
     - __output_dim__: dimension of the internal projections and the final output.
     - __init__: weight initialization function for the output cell. Can be the name of an existing function (str), or a Theano function (see: [initializations](../initializations.md)).
     - __inner_init__: weight initialization function for the inner cells.
     - __activation__: activation function for the output. Can be the name of an existing function (str), or a Theano function (see: [activations](../activations.md)).
     - __inner_activation__: activation function for the inner cells.
     - __weights__: list of numpy arrays to set as initial weights. The list should have 9 elements.
-    - __truncate_gradient__: Number of steps to use in truncated BPTT. See: [Theano "scan"](http://deeplearning.net/software/theano/library/scan.html).
     - __return_sequences__: Boolean. Whether to return the last output in the output sequence, or the full sequence.
+    - __go_backwards__: Boolean (default False). If True, rocess the input sequence backwards.
+    - __stateful__: Boolean (default False). If True, the last state for each sample at index i in a batch will be used as initial state for the sample of index i in the following batch.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
+
 
 - __References__: 
     - [On the Properties of Neural Machine Translation: Encoder–Decoder Approaches](http://www.aclweb.org/anthology/W14-4012)
@@ -107,10 +85,14 @@ Gated Recurrent Unit - Cho et al. 2014.
 ## LSTM
 
 ```python
-keras.layers.recurrent.LSTM(input_dim, output_dim=128, 
+keras.layers.recurrent.LSTM(output_dim,
         init='glorot_uniform', inner_init='orthogonal', forget_bias_init='one',
         activation='tanh', inner_activation='hard_sigmoid',
-        weights=None, truncate_gradient=-1, return_sequences=False)
+        weights=None,
+        return_sequences=False,
+        go_backwards=False,
+        stateful=False,
+        input_dim=None, input_length=None)
 ```
 
 Long-Short Term Memory unit - Hochreiter 1997.
@@ -121,10 +103,11 @@ Long-Short Term Memory unit - Hochreiter 1997.
     - if `return_sequences`: 3D tensor with shape: `(nb_samples, timesteps, output_dim)`.
     - else: 2D tensor with shape: `(nb_samples, output_dim)`.
 
-- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to true.
+- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to true. **Note:** for the time being, masking in only supported with Theano.
+
+- __Notes__: When using the TensorFlow backend, the number of timesteps used must be fixed. Make sure to pass an `input_length` int argument or a complete `input_shape` tuple argument.
 
 - __Arguments__:
-    - __input_dim__: dimension of the input.
     - __output_dim__: dimension of the internal projections and the final output.
     - __init__: weight initialization function for the output cell. Can be the name of an existing function (str), or a Theano function (see: [initializations](../initializations.md)).
     - __inner_init__: weight initialization function for the inner cells.
@@ -132,8 +115,12 @@ Long-Short Term Memory unit - Hochreiter 1997.
     - __activation__: activation function for the output. Can be the name of an existing function (str), or a Theano function (see: [activations](../activations.md)).
     - __inner_activation__: activation function for the inner cells.
     - __weights__: list of numpy arrays to set as initial weights. The list should have 12 elements.
-    - __truncate_gradient__: Number of steps to use in truncated BPTT. See: [Theano "scan"](http://deeplearning.net/software/theano/library/scan.html).
     - __return_sequences__: Boolean. Whether to return the last output in the output sequence, or the full sequence.
+    - __go_backwards__: Boolean (default False). If True, rocess the input sequence backwards.
+    - __stateful__: Boolean (default False). If True, the last state for each sample at index i in a batch will be used as initial state for the sample of index i in the following batch.
+    - __input_dim__: dimensionality of the input (integer). This argument (or alternatively, the keyword argument `input_shape`) is required when using this layer as the first layer in a model.
+    - __input_length__: Length of input sequences, when it is constant. This argument is required if you are going to connect `Flatten` then `Dense` layers upstream (without it, the shape of the dense outputs cannot be computed).
+
 
 - __References__: 
     - [Long short-term memory](http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf) (original 1997 paper)
@@ -141,39 +128,3 @@ Long-Short Term Memory unit - Hochreiter 1997.
     - [Supervised sequence labelling with recurrent neural networks](http://www.cs.toronto.edu/~graves/preprint.pdf)
 
 ---
-
-## JZS1, JZS2, JZS3
-
-```python
-keras.layers.recurrent.JZS1(input_dim, output_dim=128, 
-        init='glorot_uniform', inner_init='orthogonal', 
-        activation='tanh', inner_activation='sigmoid',
-        weights=None, truncate_gradient=-1, return_sequences=False)
-```
-
-Top 3 RNN architectures evolved from the evaluation of thousands of models. Serves as alternatives to LSTMs and GRUs. Corresponds to `MUT1`, `MUT2`, and `MUT3` architectures described in the paper: An Empirical Exploration of Recurrent Network Architectures, Jozefowicz et al. 2015.
-
-- __Input shape__: 3D tensor with shape: `(nb_samples, timesteps, input_dim)`.
-
-- __Output shape__:
-    - if `return_sequences`: 3D tensor with shape: `(nb_samples, timesteps, output_dim)`.
-    - else: 2D tensor with shape: `(nb_samples, output_dim)`.
-
-- __Masking__: This layer supports masking for input data with a variable number of timesteps To introduce masks to your data, use an [Embedding](embeddings.md) layer with the `mask_zero` parameter set to true.
-
-- __Arguments__:
-    - __input_dim__: dimension of the input.
-    - __output_dim__: dimension of the internal projections and the final output.
-    - __init__: weight initialization function for the output cell. Can be the name of an existing function (str), or a Theano function (see: [initializations](../initializations.md)).
-    - __inner_init__: weight initialization function for the inner cells.
-    - __activation__: activation function for the output. Can be the name of an existing function (str), or a Theano function (see: [activations](../activations.md)).
-    - __inner_activation__: activation function for the inner cells.
-    - __weights__: list of numpy arrays to set as initial weights. The list should have 9 elements.
-    - __truncate_gradient__: Number of steps to use in truncated BPTT. See: [Theano "scan"](http://deeplearning.net/software/theano/library/scan.html).
-    - __return_sequences__: Boolean. Whether to return the last output in the output sequence, or the full sequence.
-
-- __References__: 
-    - [An Empirical Exploration of Recurrent Network Architectures](http://www.jmlr.org/proceedings/papers/v37/jozefowicz15.pdf)
-            
-            
-                

docs/sources/models.md

@@ -12,9 +12,8 @@ model = keras.models.Sequential()
             - __optimizer__: str (name of optimizer) or optimizer object. See [optimizers](optimizers.md).
             - __loss__: str (name of objective function) or objective function. See [objectives](objectives.md).
             - __class_mode__: one of "categorical", "binary". This is only used for computing classification accuracy or using the predict_classes method.
-            - __theano_mode__: A `theano.compile.mode.Mode` ([reference](http://deeplearning.net/software/theano/library/compile/mode.html)) instance controlling specifying compilation options.
     - __fit__(X, y, batch_size=128, nb_epoch=100, verbose=1, validation_split=0., validation_data=None, shuffle=True, show_accuracy=False, callbacks=[], class_weight=None, sample_weight=None): Train a model for a fixed number of epochs.
-        - __Return__: a history dictionary with a record of training loss values at successive epochs, as well as validation loss values (if applicable), accuracy (if applicable), etc.
+        - __Return__: a history object. It `history` attribute is a record of training loss values at successive epochs, as well as validation loss values (if applicable).
         - __Arguments__:
             - __X__: data.
             - __y__: labels.
@@ -27,8 +26,8 @@ model = keras.models.Sequential()
             - __shuffle__: boolean or str (for 'batch'). Whether to shuffle the samples at each epoch. 'batch' is a special option for dealing with the limitations of HDF5 data; it shuffles in batch-sized chunks.
             - __show_accuracy__: boolean. Whether to display class accuracy in the logs to stdout at each epoch.
             - __class_weight__: dictionary mapping classes to a weight value, used for scaling the loss function (during training only).
-            - __sample_weight__: list or numpy array with 1:1 mapping to the training samples, used for scaling the loss function (during training only). For time-distributed data, there is one weight per sample *per timestep*, i.e. if your output data is shaped `(nb_samples, timesteps, output_dim)`, your mask should be of shape `(nb_samples, timesteps)`. This allows you to mask out or reweight individual output timesteps, which is useful in sequence to sequence learning.
-    - __evaluate__(X, y, batch_size=128, show_accuracy=False, verbose=1): Show performance of the model over some validation data.
+            - __sample_weight__: list or numpy array with 1:1 mapping to the training samples, used for scaling the loss function (during training only). For time-distributed data, there is one weight per sample *per timestep*, i.e. if your output data is shaped `(nb_samples, timesteps, output_dim)`, your mask should be of shape `(nb_samples, timesteps, 1)`. This allows you to mask out or reweight individual output timesteps, which is useful in sequence to sequence learning.
+    - __evaluate__(X, y, batch_size=128, show_accuracy=False, verbose=1, sample_weight=None): Show performance of the model over some validation data.
         - __Return__: The loss score over the data, or a `(loss, accuracy)` tuple if `show_accuracy=True`.
         - __Arguments__: Same meaning as fit method above. verbose is used as a binary flag (progress bar or nothing).
     - __predict__(X, batch_size=128, verbose=1):
@@ -37,9 +36,9 @@ model = keras.models.Sequential()
     - __predict_classes__(X, batch_size=128, verbose=1): Return an array of class predictions for some test data.
         - __Return__: An array of labels for some test data.
         - __Arguments__: Same meaning as fit method above. verbose is used as a binary flag (progress bar or nothing).
-    - __train_on_batch__(X, y, accuracy=False): Single gradient update on one batch.
+    - __train_on_batch__(X, y, accuracy=False, class_weight=None, sample_weight=None): Single gradient update on one batch.
         - __Return__: loss over the data, or tuple `(loss, accuracy)` if `accuracy=True`.
-    - __test_on_batch__(X, y, accuracy=False): Single performance evaluation on one batch.
+    - __test_on_batch__(X, y, accuracy=False, sample_weight=None): Single performance evaluation on one batch.
         - __Return__: loss over the data, or tuple `(loss, accuracy)` if `accuracy=True`.
     - __save_weights__(fname, overwrite=False): Store the weights of all layers to a HDF5 file. If overwrite==False and the file already exists, an exception will be thrown.
     - __load_weights__(fname): Sets the weights of a model, based to weights stored by __save_weights__. You can only __load_weights__ on a savefile from a model with an identical architecture. __load_weights__ can be called either before or after the __compile__ step.
@@ -52,7 +51,7 @@ from keras.layers.core import Dense, Dropout, Activation
 from keras.optimizers import SGD
 
 model = Sequential()
-model.add(Dense(64, 2, init='uniform'))
+model.add(Dense(2, init='uniform', input_dim=64))
 model.add(Activation('softmax'))
 
 model.compile(loss='mse', optimizer='sgd')
@@ -125,10 +124,10 @@ Arbitrary connection graph. It can have any number of inputs and outputs, with e
 model = keras.models.Graph()
 ```
 - __Methods__:
-    - __add_input__(name, ndim=2, dtype='float'): Add an input with shape dimensionality `ndim`. 
+    - __add_input__(name, input_shape, dtype='float'): Add an input with shape dimensionality `ndim`. 
         - __Arguments__:
-            - __ndim__: Use `ndim=2` for vector input `(samples, features)`, ndim=3 for temporal input `(samples, time, features)`, ndim=4 for image input `(samples, channels, height, width)`.
-            - __dtype__: `float` or `int`. Use `int` if the input is connected to an Embedding layer, `float` otherwise.
+            - __input_shape__: Integer tuple, shape of the expected input (not including the samples axis). E.g. (10,) for 10-dimensional vectors, (None, 128) for sequences (of variable length) of 128-dimensional vectors, (3, 32, 32) for 32x32 images with RGB channels.
+            - __dtype__: `float` or `int`. Type of the expected input data.
     - __add_output__(name, input=None, inputs=[], merge_mode='concat'): Add an output connect to `input` or `inputs`.
         - __Arguments__:
             - __name__: str. unique identifier of the output.
@@ -142,12 +141,19 @@ model = keras.models.Graph()
             - __input__: str name of the node/input that the node is connected to. Only specify *one* of either `input` or `inputs`.
             - __inputs__: list of str names of the node that the node is connected to.
             - __merge_mode__: "sum" or "concat". Only applicable if `inputs` list is specified. Merge mode for the different inputs.
+    - __add_shared_node__(layer, name, inputs=[], merge_mode=None, outputs=[]): Add a shared node connected to `inputs`. A shared node is a layer that will be applied separately to every incoming input, and that uses only one set of weights. The merging operation occurs on the outputs of the layer. 
+        - __Arguments__:
+            - __layer__: Layer instance.
+            - __name__: str. unique identifier of the node.
+            - __inputs__: list of str names of the node that the node is connected to.
+            - __merge_mode__: Merge mode for the different inputs.
+            - __outputs__: Optional. List of names for outputs, when merge_mode = None.
     - __compile__(optimizer, loss):
         - __Arguments__:
             - __optimizer__: str (name of optimizer) or optimizer object. See [optimizers](optimizers.md).
             - __loss__: dictionary mapping the name(s) of the output(s) to a loss function (string name of objective function or objective function. See [objectives](objectives.md)).
     - __fit__(data, batch_size=128, nb_epoch=100, verbose=1, validation_split=0., validation_data=None, shuffle=True, callbacks=[]): Train a model for a fixed number of epochs.
-        - __Return__: a history dictionary with a record of training loss values at successive epochs, as well as validation loss values (if applicable).
+        - __Return__: a history object. It `history` attribute is a record of training loss values at successive epochs, as well as validation loss values (if applicable).
         - __Arguments__:
             - __data__:dictionary mapping input names out outputs names to appropriate numpy arrays. All arrays should contain the same number of samples.
             - __batch_size__: int. Number of samples per gradient update.
@@ -176,10 +182,10 @@ __Examples__:
 ```python
 # graph model with one input and two outputs
 graph = Graph()
-graph.add_input(name='input', ndim=2)
-graph.add_node(Dense(32, 16), name='dense1', input='input')
-graph.add_node(Dense(32, 4), name='dense2', input='input')
-graph.add_node(Dense(16, 4), name='dense3', input='dense1')
+graph.add_input(name='input', input_shape=(32,))
+graph.add_node(Dense(16), name='dense1', input='input')
+graph.add_node(Dense(4), name='dense2', input='input')
+graph.add_node(Dense(4), name='dense3', input='dense1')
 graph.add_output(name='output1', input='dense2')
 graph.add_output(name='output2', input='dense3')
 
@@ -191,11 +197,11 @@ history = graph.fit({'input':X_train, 'output1':y_train, 'output2':y2_train}, nb
 ```python
 # graph model with two inputs and one output
 graph = Graph()
-graph.add_input(name='input1', ndim=2)
-graph.add_input(name='input2', ndim=2)
-graph.add_node(Dense(32, 16), name='dense1', input='input1')
-graph.add_node(Dense(32, 4), name='dense2', input='input2')
-graph.add_node(Dense(16, 4), name='dense3', input='dense1')
+graph.add_input(name='input1', input_shape=(32,))
+graph.add_input(name='input2', input_shape=(32,))
+graph.add_node(Dense(16), name='dense1', input='input1')
+graph.add_node(Dense(4), name='dense2', input='input2')
+graph.add_node(Dense(4), name='dense3', input='dense1')
 graph.add_output(name='output', inputs=['dense2', 'dense3'], merge_mode='sum')
 graph.compile('rmsprop', {'output':'mse'})
 

docs/sources/objectives.md

@@ -19,6 +19,7 @@ For a few examples of such functions, check out the [objectives source](https://
 ## Available objectives
 
 - __mean_squared_error__ / __mse__
+- __root_mean_squared_error__ / __rmse__
 - __mean_absolute_error__ / __mae__
 - __mean_absolute_percentage_error__ / __mape__
 - __mean_squared_logarithmic_error__ / __msle__

docs/sources/optimizers.md

@@ -5,7 +5,7 @@ An optimizer is one of the two arguments required for compiling a Keras model:
 
 ```python
 model = Sequential()
-model.add(Dense(20, 64, init='uniform'))
+model.add(Dense(64, init='uniform', input_dim=10))
 model.add(Activation('tanh'))
 model.add(Activation('softmax'))
 

docs/sources/regularizers.md

@@ -15,7 +15,7 @@ These layers expose 3 keyword arguments:
 
 ```python
 from keras.regularizers import l2, activity_l2
-model.add(Dense(64, 64, W_regularizer=l2(0.01), activity_regularizer=activity_l2(0.01)))
+model.add(Dense(64, input_dim=64, W_regularizer=l2(0.01), activity_regularizer=activity_l2(0.01)))
 ```
 
 ## Available penalties

docs/sources/visualization.md

@@ -0,0 +1,20 @@
+
+## Model visualization
+
+The `keras.utils.visualize_util` module provides utility functions to plot
+a Keras model (using graphviz).
+
+This will plot a graph of the model and save it to a file:
+```python
+from keras.utils.visualize_util import plot
+plot(model, to_file='model.png')
+```
+
+You can also directly obtain the `pydot.Graph` object and render it yourself,
+for example to show it in an ipython notebook :
+```python
+from IPython.display import SVG
+from keras.utils.visualize_util import to_graph
+
+SVG(to_graph(model).create(prog='dot', format='svg'))
+```

examples/addition_rnn.py

@@ -1,10 +1,10 @@
 # -*- coding: utf-8 -*-
 from __future__ import print_function
 from keras.models import Sequential, slice_X
-from keras.layers.core import Activation, Dense, RepeatVector
+from keras.layers.core import Activation, TimeDistributedDense, RepeatVector
 from keras.layers import recurrent
-from sklearn.utils import shuffle
 import numpy as np
+from six.moves import range
 
 """
 An implementation of sequence to sequence learning for performing addition
@@ -12,11 +12,6 @@ Input: "535+61"
 Output: "596"
 Padding is handled by using a repeated sentinel character (space)
 
-By default, the JZS1 recurrent neural network is used
-JZS1 was an "evolved" recurrent neural network performing well on arithmetic benchmark in:
-"An Empirical Exploration of Recurrent Network Architectures"
-http://jmlr.org/proceedings/papers/v37/jozefowicz15.pdf
-
 Input may optionally be inverted, shown to increase performance in many tasks in:
 "Learning to Execute"
 http://arxiv.org/abs/1410.4615
@@ -25,20 +20,17 @@ and
 http://papers.nips.cc/paper/5346-sequence-to-sequence-learning-with-neural-networks.pdf
 Theoretically it introduces shorter term dependencies between source and target.
 
-
 Two digits inverted:
-+ One layer JZS1 (128 HN), 5k training examples = 99% train/test accuracy in 55 epochs
++ One layer LSTM (128 HN), 5k training examples = 99% train/test accuracy in 55 epochs
 
 Three digits inverted:
-+ One layer JZS1 (128 HN), 50k training examples = 99% train/test accuracy in 100 epochs
-
++ One layer LSTM (128 HN), 50k training examples = 99% train/test accuracy in 100 epochs
 
 Four digits inverted:
-+ One layer JZS1 (128 HN), 400k training examples = 99% train/test accuracy in 20 epochs
-
++ One layer LSTM (128 HN), 400k training examples = 99% train/test accuracy in 20 epochs
 
 Five digits inverted:
-+ One layer JZS1 (128 HN), 550k training examples = 99% train/test accuracy in 30 epochs
++ One layer LSTM (128 HN), 550k training examples = 99% train/test accuracy in 30 epochs
 
 """
 
@@ -78,8 +70,8 @@ class colors:
 TRAINING_SIZE = 50000
 DIGITS = 3
 INVERT = True
-# Try replacing JZS1 with LSTM, GRU, or SimpleRNN
-RNN = recurrent.JZS1
+# Try replacing GRU, or SimpleRNN
+RNN = recurrent.LSTM
 HIDDEN_SIZE = 128
 BATCH_SIZE = 128
 LAYERS = 1
@@ -93,7 +85,7 @@ expected = []
 seen = set()
 print('Generating data...')
 while len(questions) < TRAINING_SIZE:
-    f = lambda: int(''.join(np.random.choice(list('0123456789')) for i in xrange(np.random.randint(1, DIGITS + 1))))
+    f = lambda: int(''.join(np.random.choice(list('0123456789')) for i in range(np.random.randint(1, DIGITS + 1))))
     a, b = f(), f()
     # Skip any addition questions we've already seen
     # Also skip any such that X+Y == Y+X (hence the sorting)
@@ -122,23 +114,33 @@ for i, sentence in enumerate(expected):
     y[i] = ctable.encode(sentence, maxlen=DIGITS + 1)
 
 # Shuffle (X, y) in unison as the later parts of X will almost all be larger digits
-X, y = shuffle(X, y)
+indices = np.arange(len(y))
+np.random.shuffle(indices)
+X = X[indices]
+y = y[indices]
+
 # Explicitly set apart 10% for validation data that we never train over
 split_at = len(X) - len(X) / 10
 (X_train, X_val) = (slice_X(X, 0, split_at), slice_X(X, split_at))
 (y_train, y_val) = (y[:split_at], y[split_at:])
 
+print(X_train.shape)
+print(y_train.shape)
+
 print('Build model...')
 model = Sequential()
 # "Encode" the input sequence using an RNN, producing an output of HIDDEN_SIZE
-model.add(RNN(len(chars), HIDDEN_SIZE))
+# note: in a situation where your input sequences have a variable length,
+# use input_shape=(None, nb_feature).
+model.add(RNN(HIDDEN_SIZE, input_shape=(MAXLEN, len(chars))))
 # For the decoder's input, we repeat the encoded input for each time step
 model.add(RepeatVector(DIGITS + 1))
 # The decoder RNN could be multiple layers stacked or a single layer
-for _ in xrange(LAYERS):
-    model.add(RNN(HIDDEN_SIZE, HIDDEN_SIZE, return_sequences=True))
+for _ in range(LAYERS):
+    model.add(RNN(HIDDEN_SIZE, return_sequences=True))
+
 # For each of step of the output sequence, decide which character should be chosen
-model.add(Dense(HIDDEN_SIZE, len(chars)))
+model.add(TimeDistributedDense(len(chars)))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='adam')
@@ -148,10 +150,10 @@ for iteration in range(1, 200):
     print()
     print('-' * 50)
     print('Iteration', iteration)
-    model.fit(X, y, batch_size=BATCH_SIZE, nb_epoch=1, validation_data=(X_val, y_val), show_accuracy=True)
+    model.fit(X_train, y_train, batch_size=BATCH_SIZE, nb_epoch=1, validation_data=(X_val, y_val), show_accuracy=True)
     ###
     # Select 10 samples from the validation set at random so we can visualize errors
-    for i in xrange(10):
+    for i in range(10):
         ind = np.random.randint(0, len(X_val))
         rowX, rowy = X_val[np.array([ind])], y_val[np.array([ind])]
         preds = model.predict_classes(rowX, verbose=0)

examples/babi_memnn.py

@@ -0,0 +1,203 @@
+from __future__ import print_function
+from keras.models import Sequential
+from keras.layers.embeddings import Embedding
+from keras.layers.core import Activation, Dense, Merge, Permute, Dropout
+from keras.layers.recurrent import LSTM
+from keras.datasets.data_utils import get_file
+from keras.preprocessing.sequence import pad_sequences
+from functools import reduce
+import tarfile
+import numpy as np
+import re
+
+"""
+Train a memory network on the bAbI dataset.
+
+References:
+- Jason Weston, Antoine Bordes, Sumit Chopra, Tomas Mikolov, Alexander M. Rush,
+  "Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks",
+  http://arxiv.org/abs/1503.08895
+
+- Sainbayar Sukhbaatar, Arthur Szlam, Jason Weston, Rob Fergus,
+  "End-To-End Memory Networks",
+  http://arxiv.org/abs/1503.08895
+
+Reaches 93% accuracy on task 'single_supporting_fact_10k' after 70 epochs.
+Time per epoch: 3s on CPU (core i7).
+"""
+
+
+def tokenize(sent):
+    '''Return the tokens of a sentence including punctuation.
+
+    >>> tokenize('Bob dropped the apple. Where is the apple?')
+    ['Bob', 'dropped', 'the', 'apple', '.', 'Where', 'is', 'the', 'apple', '?']
+    '''
+    return [x.strip() for x in re.split('(\W+)?', sent) if x.strip()]
+
+
+def parse_stories(lines, only_supporting=False):
+    '''Parse stories provided in the bAbi tasks format
+
+    If only_supporting is true, only the sentences that support the answer are kept.
+    '''
+    data = []
+    story = []
+    for line in lines:
+        line = line.decode('utf-8').strip()
+        nid, line = line.split(' ', 1)
+        nid = int(nid)
+        if nid == 1:
+            story = []
+        if '\t' in line:
+            q, a, supporting = line.split('\t')
+            q = tokenize(q)
+            substory = None
+            if only_supporting:
+                # Only select the related substory
+                supporting = map(int, supporting.split())
+                substory = [story[i - 1] for i in supporting]
+            else:
+                # Provide all the substories
+                substory = [x for x in story if x]
+            data.append((substory, q, a))
+            story.append('')
+        else:
+            sent = tokenize(line)
+            story.append(sent)
+    return data
+
+
+def get_stories(f, only_supporting=False, max_length=None):
+    '''Given a file name, read the file, retrieve the stories, and then convert the sentences into a single story.
+
+    If max_length is supplied, any stories longer than max_length tokens will be discarded.
+    '''
+    data = parse_stories(f.readlines(), only_supporting=only_supporting)
+    flatten = lambda data: reduce(lambda x, y: x + y, data)
+    data = [(flatten(story), q, answer) for story, q, answer in data if not max_length or len(flatten(story)) < max_length]
+    return data
+
+
+def vectorize_stories(data, word_idx, story_maxlen, query_maxlen):
+    X = []
+    Xq = []
+    Y = []
+    for story, query, answer in data:
+        x = [word_idx[w] for w in story]
+        xq = [word_idx[w] for w in query]
+        y = np.zeros(len(word_idx) + 1)  # let's not forget that index 0 is reserved
+        y[word_idx[answer]] = 1
+        X.append(x)
+        Xq.append(xq)
+        Y.append(y)
+    return (pad_sequences(X, maxlen=story_maxlen),
+            pad_sequences(Xq, maxlen=query_maxlen), np.array(Y))
+
+
+path = get_file('babi-tasks-v1-2.tar.gz',
+                origin='http://www.thespermwhale.com/jaseweston/babi/tasks_1-20_v1-2.tar.gz')
+tar = tarfile.open(path)
+
+challenges = {
+    # QA1 with 10,000 samples
+    'single_supporting_fact_10k': 'tasks_1-20_v1-2/en-10k/qa1_single-supporting-fact_{}.txt',
+    # QA2 with 10,000 samples
+    'two_supporting_facts_10k': 'tasks_1-20_v1-2/en-10k/qa2_two-supporting-facts_{}.txt',
+}
+challenge_type = 'single_supporting_fact_10k'
+challenge = challenges[challenge_type]
+
+print('Extracting stories for the challenge:', challenge_type)
+train_stories = get_stories(tar.extractfile(challenge.format('train')))
+test_stories = get_stories(tar.extractfile(challenge.format('test')))
+
+vocab = sorted(reduce(lambda x, y: x | y, (set(story + q + [answer]) for story, q, answer in train_stories + test_stories)))
+# Reserve 0 for masking via pad_sequences
+vocab_size = len(vocab) + 1
+story_maxlen = max(map(len, (x for x, _, _ in train_stories + test_stories)))
+query_maxlen = max(map(len, (x for _, x, _ in train_stories + test_stories)))
+
+print('-')
+print('Vocab size:', vocab_size, 'unique words')
+print('Story max length:', story_maxlen, 'words')
+print('Query max length:', query_maxlen, 'words')
+print('Number of training stories:', len(train_stories))
+print('Number of test stories:', len(test_stories))
+print('-')
+print('Here\'s what a "story" tuple looks like (input, query, answer):')
+print(train_stories[0])
+print('-')
+print('Vectorizing the word sequences...')
+
+word_idx = dict((c, i + 1) for i, c in enumerate(vocab))
+inputs_train, queries_train, answers_train = vectorize_stories(train_stories, word_idx, story_maxlen, query_maxlen)
+inputs_test, queries_test, answers_test = vectorize_stories(test_stories, word_idx, story_maxlen, query_maxlen)
+
+print('-')
+print('inputs: integer tensor of shape (samples, max_length)')
+print('inputs_train shape:', inputs_train.shape)
+print('inputs_test shape:', inputs_test.shape)
+print('-')
+print('queries: integer tensor of shape (samples, max_length)')
+print('queries_train shape:', queries_train.shape)
+print('queries_test shape:', queries_test.shape)
+print('-')
+print('answers: binary (1 or 0) tensor of shape (samples, vocab_size)')
+print('answers_train shape:', answers_train.shape)
+print('answers_test shape:', answers_test.shape)
+print('-')
+print('Compiling...')
+
+# embed the input sequence into a sequence of vectors
+input_encoder_m = Sequential()
+input_encoder_m.add(Embedding(input_dim=vocab_size,
+                              output_dim=64,
+                              input_length=story_maxlen))
+# output: (samples, story_maxlen, embedding_dim)
+# embed the question into a sequence of vectors
+question_encoder = Sequential()
+question_encoder.add(Embedding(input_dim=vocab_size,
+                               output_dim=64,
+                               input_length=query_maxlen))
+# output: (samples, query_maxlen, embedding_dim)
+# compute a 'match' between input sequence elements (which are vectors)
+# and the question vector sequence
+match = Sequential()
+match.add(Merge([input_encoder_m, question_encoder],
+                mode='dot',
+                dot_axes=[(2,), (2,)]))
+# output: (samples, story_maxlen, query_maxlen)
+# embed the input into a single vector with size = story_maxlen:
+input_encoder_c = Sequential()
+input_encoder_c.add(Embedding(input_dim=vocab_size,
+                              output_dim=query_maxlen,
+                              input_length=story_maxlen))
+# output: (samples, story_maxlen, query_maxlen)
+# sum the match vector with the input vector:
+response = Sequential()
+response.add(Merge([match, input_encoder_c], mode='sum'))
+# output: (samples, story_maxlen, query_maxlen)
+response.add(Permute((2, 1)))  # output: (samples, query_maxlen, story_maxlen)
+
+# concatenate the match vector with the question vector,
+# and do logistic regression on top
+answer = Sequential()
+answer.add(Merge([response, question_encoder], mode='concat', concat_axis=-1))
+# the original paper uses a matrix multiplication for this reduction step.
+# we choose to use a RNN instead.
+answer.add(LSTM(64))
+# one regularization layer -- more would probably be needed.
+answer.add(Dropout(0.25))
+answer.add(Dense(vocab_size))
+# we output a probability distribution over the vocabulary
+answer.add(Activation('softmax'))
+
+answer.compile(optimizer='rmsprop', loss='categorical_crossentropy')
+# Note: you could use a Graph model to avoid repeat the input twice
+answer.fit([inputs_train, queries_train, inputs_train], answers_train,
+           batch_size=32,
+           nb_epoch=70,
+           show_accuracy=True,
+           validation_data=([inputs_test, queries_test, inputs_test], answers_test))
+

examples/babi_rnn.py

@@ -126,14 +126,14 @@ def get_stories(f, only_supporting=False, max_length=None):
     return data
 
 
-def vectorize_stories(data):
+def vectorize_stories(data, word_idx, story_maxlen, query_maxlen):
     X = []
     Xq = []
     Y = []
     for story, query, answer in data:
         x = [word_idx[w] for w in story]
         xq = [word_idx[w] for w in query]
-        y = np.zeros(vocab_size)
+        y = np.zeros(len(word_idx) + 1)  # let's not forget that index 0 is reserved
         y[word_idx[answer]] = 1
         X.append(x)
         Xq.append(xq)
@@ -168,8 +168,8 @@ word_idx = dict((c, i + 1) for i, c in enumerate(vocab))
 story_maxlen = max(map(len, (x for x, _, _ in train + test)))
 query_maxlen = max(map(len, (x for _, x, _ in train + test)))
 
-X, Xq, Y = vectorize_stories(train)
-tX, tXq, tY = vectorize_stories(test)
+X, Xq, Y = vectorize_stories(train, word_idx, story_maxlen, query_maxlen)
+tX, tXq, tY = vectorize_stories(test, word_idx, story_maxlen, query_maxlen)
 
 print('vocab = {}'.format(vocab))
 print('X.shape = {}'.format(X.shape))
@@ -181,15 +181,15 @@ print('Build model...')
 
 sentrnn = Sequential()
 sentrnn.add(Embedding(vocab_size, EMBED_HIDDEN_SIZE, mask_zero=True))
-sentrnn.add(RNN(EMBED_HIDDEN_SIZE, SENT_HIDDEN_SIZE, return_sequences=False))
+sentrnn.add(RNN(SENT_HIDDEN_SIZE, return_sequences=False))
 
 qrnn = Sequential()
 qrnn.add(Embedding(vocab_size, EMBED_HIDDEN_SIZE))
-qrnn.add(RNN(EMBED_HIDDEN_SIZE, QUERY_HIDDEN_SIZE, return_sequences=False))
+qrnn.add(RNN(QUERY_HIDDEN_SIZE, return_sequences=False))
 
 model = Sequential()
 model.add(Merge([sentrnn, qrnn], mode='concat'))
-model.add(Dense(SENT_HIDDEN_SIZE + QUERY_HIDDEN_SIZE, vocab_size, activation='softmax'))
+model.add(Dense(vocab_size, activation='softmax'))
 
 model.compile(optimizer='adam', loss='categorical_crossentropy', class_mode='categorical')
 

examples/cifar10_cnn.py

@@ -28,7 +28,12 @@ nb_classes = 10
 nb_epoch = 200
 data_augmentation = True
 
-# the data, shuffled and split between tran and test sets
+# input image dimensions
+img_rows, img_cols = 32, 32
+# the CIFAR10 images are RGB
+img_channels = 3
+
+# the data, shuffled and split between train and test sets
 (X_train, y_train), (X_test, y_test) = cifar10.load_data()
 print('X_train shape:', X_train.shape)
 print(X_train.shape[0], 'train samples')
@@ -40,39 +45,39 @@ Y_test = np_utils.to_categorical(y_test, nb_classes)
 
 model = Sequential()
 
-model.add(Convolution2D(32, 3, 3, 3, border_mode='full'))
+model.add(Convolution2D(32, 3, 3, border_mode='same',
+                        input_shape=(img_channels, img_rows, img_cols)))
 model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
+model.add(Convolution2D(32, 3, 3))
 model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+model.add(MaxPooling2D(pool_size=(2, 2)))
 model.add(Dropout(0.25))
 
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full'))
+model.add(Convolution2D(64, 3, 3, border_mode='same'))
 model.add(Activation('relu'))
-model.add(Convolution2D(64, 64, 3, 3))
+model.add(Convolution2D(64, 3, 3))
 model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+model.add(MaxPooling2D(pool_size=(2, 2)))
 model.add(Dropout(0.25))
 
 model.add(Flatten())
-model.add(Dense(64*8*8, 512))
+model.add(Dense(512))
 model.add(Activation('relu'))
 model.add(Dropout(0.5))
-
-model.add(Dense(512, nb_classes))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 
 # let's train the model using SGD + momentum (how original).
 sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
 model.compile(loss='categorical_crossentropy', optimizer=sgd)
 
+X_train = X_train.astype("float32")
+X_test = X_test.astype("float32")
+X_train /= 255
+X_test /= 255
+
 if not data_augmentation:
     print("Not using data augmentation or normalization")
-
-    X_train = X_train.astype("float32")
-    X_test = X_test.astype("float32")
-    X_train /= 255
-    X_test /= 255
     model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch)
     score = model.evaluate(X_test, Y_test, batch_size=batch_size)
     print('Test score:', score)

examples/imdb_bidirectional_lstm.py

@@ -0,0 +1,63 @@
+from __future__ import absolute_import
+from __future__ import print_function
+import numpy as np
+np.random.seed(1337)  # for reproducibility
+
+from keras.preprocessing import sequence
+from keras.utils.np_utils import accuracy
+from keras.models import Graph
+from keras.layers.core import Dense, Dropout
+from keras.layers.embeddings import Embedding
+from keras.layers.recurrent import LSTM
+from keras.datasets import imdb
+
+'''
+    Train a Bidirectional LSTM on the IMDB sentiment classification task.
+
+    GPU command:
+        THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python imdb_bidirectional_lstm.py
+
+    Output after 4 epochs on CPU: ~0.8146
+    Time per epoch on CPU (Core i7): ~150s.
+'''
+
+max_features = 20000
+maxlen = 100  # cut texts after this number of words (among top max_features most common words)
+batch_size = 32
+
+print("Loading data...")
+(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features,
+                                                      test_split=0.2)
+print(len(X_train), 'train sequences')
+print(len(X_test), 'test sequences')
+
+print("Pad sequences (samples x time)")
+X_train = sequence.pad_sequences(X_train, maxlen=maxlen)
+X_test = sequence.pad_sequences(X_test, maxlen=maxlen)
+print('X_train shape:', X_train.shape)
+print('X_test shape:', X_test.shape)
+y_train = np.array(y_train)
+y_test = np.array(y_test)
+
+print('Build model...')
+model = Graph()
+model.add_input(name='input', input_shape=(maxlen,), dtype=int)
+model.add_node(Embedding(max_features, 128, input_length=maxlen),
+               name='embedding', input='input')
+model.add_node(LSTM(64), name='forward', input='embedding')
+model.add_node(LSTM(64, go_backwards=True), name='backward', input='embedding')
+model.add_node(Dropout(0.5), name='dropout', inputs=['forward', 'backward'])
+model.add_node(Dense(1, activation='sigmoid'), name='sigmoid', input='dropout')
+model.add_output(name='output', input='sigmoid')
+
+# try using different optimizers and different optimizer configs
+model.compile('adam', {'output': 'binary_crossentropy'})
+
+print("Train...")
+model.fit({'input': X_train, 'output': y_train},
+          batch_size=batch_size,
+          nb_epoch=4)
+acc = accuracy(y_test,
+               np.round(np.array(model.predict({'input': X_test},
+                                               batch_size=batch_size)['output'])))
+print('Test accuracy:', acc)

examples/imdb_cnn.py

@@ -1,10 +1,9 @@
 from __future__ import absolute_import
 from __future__ import print_function
 import numpy as np
-np.random.seed(1337) # for reproducibility
+np.random.seed(1337)  # for reproducibility
 
 from keras.preprocessing import sequence
-from keras.optimizers import RMSprop
 from keras.models import Sequential
 from keras.layers.core import Dense, Dropout, Activation, Flatten
 from keras.layers.embeddings import Embedding
@@ -17,7 +16,7 @@ from keras.datasets import imdb
 
     Run on GPU: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python imdb_cnn.py
 
-    Get to 0.8330 test accuracy after 3 epochs. 100s/epoch on K520 GPU.
+    Get to 0.835 test accuracy after 2 epochs. 100s/epoch on K520 GPU.
 '''
 
 # set parameters:
@@ -25,10 +24,10 @@ max_features = 5000
 maxlen = 100
 batch_size = 32
 embedding_dims = 100
-nb_filters = 250
+nb_filter = 250
 filter_length = 3
 hidden_dims = 250
-nb_epoch = 3
+nb_epoch = 2
 
 print("Loading data...")
 (X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features,
@@ -47,36 +46,35 @@ model = Sequential()
 
 # we start off with an efficient embedding layer which maps
 # our vocab indices into embedding_dims dimensions
-model.add(Embedding(max_features, embedding_dims))
+model.add(Embedding(max_features, embedding_dims, input_length=maxlen))
 model.add(Dropout(0.25))
 
-# we add a Convolution1D, which will learn nb_filters
+# we add a Convolution1D, which will learn nb_filter
 # word group filters of size filter_length:
-model.add(Convolution1D(input_dim=embedding_dims,
-                        nb_filter=nb_filters,
+model.add(Convolution1D(nb_filter=nb_filter,
                         filter_length=filter_length,
                         border_mode="valid",
                         activation="relu",
                         subsample_length=1))
-
 # we use standard max pooling (halving the output of the previous layer):
 model.add(MaxPooling1D(pool_length=2))
 
-# We flatten the output of the conv layer, so that we can add a vanilla dense layer:
+# We flatten the output of the conv layer,
+# so that we can add a vanilla dense layer:
 model.add(Flatten())
 
-# Computing the output shape of a conv layer can be tricky;
-# for a good tutorial, see: http://cs231n.github.io/convolutional-networks/
-output_size = nb_filters * (((maxlen - filter_length) / 1) + 1) / 2
-
 # We add a vanilla hidden layer:
-model.add(Dense(output_size, hidden_dims))
+model.add(Dense(hidden_dims))
 model.add(Dropout(0.25))
 model.add(Activation('relu'))
 
 # We project onto a single unit output layer, and squash it with a sigmoid:
-model.add(Dense(hidden_dims, 1))
+model.add(Dense(1))
 model.add(Activation('sigmoid'))
 
-model.compile(loss='binary_crossentropy', optimizer='rmsprop', class_mode="binary")
-model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, validation_data=(X_test, y_test))
+model.compile(loss='binary_crossentropy',
+              optimizer='rmsprop',
+              class_mode="binary")
+model.fit(X_train, y_train, batch_size=batch_size,
+          nb_epoch=nb_epoch, show_accuracy=True,
+          validation_data=(X_test, y_test))

examples/imdb_cnn_lstm.py

@@ -0,0 +1,84 @@
+from __future__ import absolute_import
+from __future__ import print_function
+import numpy as np
+np.random.seed(1337)  # for reproducibility
+
+from keras.preprocessing import sequence
+from keras.optimizers import SGD, RMSprop, Adagrad
+from keras.utils import np_utils
+from keras.models import Sequential
+from keras.layers.core import Dense, Dropout, Activation
+from keras.layers.embeddings import Embedding
+from keras.layers.recurrent import LSTM, GRU, SimpleRNN
+from keras.layers.convolutional import Convolution1D, MaxPooling1D
+from keras.datasets import imdb
+
+'''
+    Train a recurrent convolutional network on the IMDB sentiment classification task.
+
+    GPU command:
+        THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python imdb_lstm.py
+
+    Get to 0.8498 test accuracy after 2 epochs. 41s/epoch on K520 GPU.
+'''
+
+# Embedding
+max_features = 20000
+maxlen = 100
+embedding_size = 128
+
+# Convolution
+filter_length = 3
+nb_filter = 64
+pool_length = 2
+
+# LSTM
+lstm_output_size = 70
+
+# Training
+batch_size = 30
+nb_epoch = 2
+
+'''
+Note:
+batch_size is highly sensitive.
+Only 2 epochs are needed as the dataset is very small.
+'''
+
+print("Loading data...")
+(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features, test_split=0.2)
+print(len(X_train), 'train sequences')
+print(len(X_test), 'test sequences')
+
+print("Pad sequences (samples x time)")
+X_train = sequence.pad_sequences(X_train, maxlen=maxlen)
+X_test = sequence.pad_sequences(X_test, maxlen=maxlen)
+print('X_train shape:', X_train.shape)
+print('X_test shape:', X_test.shape)
+
+print('Build model...')
+
+model = Sequential()
+model.add(Embedding(max_features, embedding_size, input_length=maxlen))
+model.add(Dropout(0.25))
+model.add(Convolution1D(nb_filter=nb_filter,
+                        filter_length=filter_length,
+                        border_mode="valid",
+                        activation="relu",
+                        subsample_length=1))
+model.add(MaxPooling1D(pool_length=pool_length))
+model.add(LSTM(lstm_output_size))
+model.add(Dense(1))
+model.add(Activation('sigmoid'))
+
+model.compile(loss='binary_crossentropy',
+              optimizer='adam',
+              class_mode="binary")
+
+print("Train...")
+model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=nb_epoch,
+          validation_data=(X_test, y_test), show_accuracy=True)
+score, acc = model.evaluate(X_test, y_test, batch_size=batch_size,
+                            show_accuracy=True)
+print('Test score:', score)
+print('Test accuracy:', acc)

examples/imdb_lstm.py

@@ -4,12 +4,11 @@ import numpy as np
 np.random.seed(1337)  # for reproducibility
 
 from keras.preprocessing import sequence
-from keras.optimizers import SGD, RMSprop, Adagrad
 from keras.utils import np_utils
 from keras.models import Sequential
 from keras.layers.core import Dense, Dropout, Activation
 from keras.layers.embeddings import Embedding
-from keras.layers.recurrent import LSTM, GRU
+from keras.layers.recurrent import LSTM
 from keras.datasets import imdb
 
 '''
@@ -36,7 +35,8 @@ maxlen = 100  # cut texts after this number of words (among top max_features mos
 batch_size = 32
 
 print("Loading data...")
-(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features, test_split=0.2)
+(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features,
+                                                      test_split=0.2)
 print(len(X_train), 'train sequences')
 print(len(X_test), 'test sequences')
 
@@ -48,17 +48,22 @@ print('X_test shape:', X_test.shape)
 
 print('Build model...')
 model = Sequential()
-model.add(Embedding(max_features, 128))
-model.add(LSTM(128, 128))  # try using a GRU instead, for fun
+model.add(Embedding(max_features, 128, input_length=maxlen))
+model.add(LSTM(128))  # try using a GRU instead, for fun
 model.add(Dropout(0.5))
-model.add(Dense(128, 1))
+model.add(Dense(1))
 model.add(Activation('sigmoid'))
 
 # try using different optimizers and different optimizer configs
-model.compile(loss='binary_crossentropy', optimizer='adam', class_mode="binary")
+model.compile(loss='binary_crossentropy',
+              optimizer='adam',
+              class_mode="binary")
 
 print("Train...")
-model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=4, validation_data=(X_test, y_test), show_accuracy=True)
-score, acc = model.evaluate(X_test, y_test, batch_size=batch_size, show_accuracy=True)
+model.fit(X_train, y_train, batch_size=batch_size, nb_epoch=3,
+          validation_data=(X_test, y_test), show_accuracy=True)
+score, acc = model.evaluate(X_test, y_test,
+                            batch_size=batch_size,
+                            show_accuracy=True)
 print('Test score:', score)
 print('Test accuracy:', acc)

examples/kaggle_otto_nn.py

@@ -20,11 +20,11 @@ from sklearn.preprocessing import StandardScaler
 
     Compatible Python 2.7-3.4. Requires Scikit-Learn and Pandas.
 
-    Recommended to run on GPU: 
+    Recommended to run on GPU:
         Command: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python kaggle_otto_nn.py
         On EC2 g2.2xlarge instance: 19s/epoch. 6-7 minutes total training time.
 
-    Best validation score at epoch 21: 0.4881 
+    Best validation score at epoch 21: 0.4881
 
     Try it at home:
         - with/without BatchNormalization (BatchNormalization helps!)
@@ -78,7 +78,6 @@ def make_submission(y_prob, ids, encoder, fname):
             f.write('\n')
     print("Wrote submission to file {}.".format(fname))
 
-
 print("Loading data...")
 X, labels = load_data('train.csv', train=True)
 X, scaler = preprocess_data(X)
@@ -96,31 +95,29 @@ print(dims, 'dims')
 print("Building model...")
 
 model = Sequential()
-model.add(Dense(dims, 512, init='glorot_uniform'))
-model.add(PReLU((512,)))
-model.add(BatchNormalization((512,)))
+model.add(Dense(512, input_shape=(dims,)))
+model.add(PReLU())
+model.add(BatchNormalization())
 model.add(Dropout(0.5))
 
-model.add(Dense(512, 512, init='glorot_uniform'))
-model.add(PReLU((512,)))
-model.add(BatchNormalization((512,)))
+model.add(Dense(512))
+model.add(PReLU())
+model.add(BatchNormalization())
 model.add(Dropout(0.5))
 
-model.add(Dense(512, 512, init='glorot_uniform'))
-model.add(PReLU((512,)))
-model.add(BatchNormalization((512,)))
+model.add(Dense(512))
+model.add(PReLU())
+model.add(BatchNormalization())
 model.add(Dropout(0.5))
 
-model.add(Dense(512, nb_classes, init='glorot_uniform'))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer="adam")
 
 print("Training model...")
-
 model.fit(X, y, nb_epoch=20, batch_size=128, validation_split=0.15)
 
 print("Generating submission...")
-
 proba = model.predict_proba(X_test)
 make_submission(proba, ids, encoder, fname='keras-otto.csv')

examples/lstm_text_generation.py

@@ -4,7 +4,8 @@ from keras.layers.core import Dense, Activation, Dropout
 from keras.layers.recurrent import LSTM
 from keras.datasets.data_utils import get_file
 import numpy as np
-import random, sys
+import random
+import sys
 
 '''
     Example script to generate text from Nietzsche's writings.
@@ -15,7 +16,7 @@ import random, sys
     It is recommended to run this script on GPU, as recurrent
     networks are quite computationally intensive.
 
-    If you try this script on new data, make sure your corpus 
+    If you try this script on new data, make sure your corpus
     has at least ~100k characters. ~1M is better.
 '''
 
@@ -34,7 +35,7 @@ step = 3
 sentences = []
 next_chars = []
 for i in range(0, len(text) - maxlen, step):
-    sentences.append(text[i : i + maxlen])
+    sentences.append(text[i: i + maxlen])
     next_chars.append(text[i + maxlen])
 print('nb sequences:', len(sentences))
 
@@ -50,20 +51,21 @@ for i, sentence in enumerate(sentences):
 # build the model: 2 stacked LSTM
 print('Build model...')
 model = Sequential()
-model.add(LSTM(len(chars), 512, return_sequences=True))
+model.add(LSTM(512, return_sequences=True, input_shape=(maxlen, len(chars))))
 model.add(Dropout(0.2))
-model.add(LSTM(512, 512, return_sequences=False))
+model.add(LSTM(512, return_sequences=False))
 model.add(Dropout(0.2))
-model.add(Dense(512, len(chars)))
+model.add(Dense(len(chars)))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
 
-# helper function to sample an index from a probability array
+
 def sample(a, temperature=1.0):
-    a = np.log(a)/temperature
-    a = np.exp(a)/np.sum(np.exp(a))
-    return np.argmax(np.random.multinomial(1,a,1))
+    # helper function to sample an index from a probability array
+    a = np.log(a) / temperature
+    a = np.exp(a) / np.sum(np.exp(a))
+    return np.argmax(np.random.multinomial(1, a, 1))
 
 # train the model, output generated text after each iteration
 for iteration in range(1, 60):
@@ -79,7 +81,7 @@ for iteration in range(1, 60):
         print('----- diversity:', diversity)
 
         generated = ''
-        sentence = text[start_index : start_index + maxlen]
+        sentence = text[start_index: start_index + maxlen]
         generated += sentence
         print('----- Generating with seed: "' + sentence + '"')
         sys.stdout.write(generated)

examples/mnist_cnn.py

@@ -22,11 +22,20 @@ batch_size = 128
 nb_classes = 10
 nb_epoch = 12
 
+# input image dimensions
+img_rows, img_cols = 28, 28
+# number of convolutional filters to use
+nb_filters = 32
+# size of pooling area for max pooling
+nb_pool = 2
+# convolution kernel size
+nb_conv = 3
+
 # the data, shuffled and split between tran and test sets
 (X_train, y_train), (X_test, y_test) = mnist.load_data()
 
-X_train = X_train.reshape(X_train.shape[0], 1, 28, 28)
-X_test = X_test.reshape(X_test.shape[0], 1, 28, 28)
+X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)
+X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)
 X_train = X_train.astype("float32")
 X_test = X_test.astype("float32")
 X_train /= 255
@@ -41,19 +50,20 @@ Y_test = np_utils.to_categorical(y_test, nb_classes)
 
 model = Sequential()
 
-model.add(Convolution2D(32, 1, 3, 3, border_mode='full'))
+model.add(Convolution2D(nb_filters, nb_conv, nb_conv,
+                        border_mode='same',
+                        input_shape=(1, img_rows, img_cols)))
 model.add(Activation('relu'))
-model.add(Convolution2D(32, 32, 3, 3))
+model.add(Convolution2D(nb_filters, nb_conv, nb_conv))
 model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
+model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))
 model.add(Dropout(0.25))
 
 model.add(Flatten())
-model.add(Dense(32*196, 128))
+model.add(Dense(128))
 model.add(Activation('relu'))
 model.add(Dropout(0.5))
-
-model.add(Dense(128, nb_classes))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='adadelta')

examples/mnist_irnn.py

@@ -23,8 +23,8 @@ from keras.utils import np_utils
     Optimizer is replaced with RMSprop which yields more stable and steady
     improvement.
 
-    Reaches 0.93 train/test accuracy after 900 epochs (which roughly corresponds
-    to 1687500 steps in the original paper.)
+    Reaches 0.93 train/test accuracy after 900 epochs
+    (which roughly corresponds to 1687500 steps in the original paper.)
 '''
 
 batch_size = 32
@@ -34,7 +34,6 @@ hidden_units = 100
 
 learning_rate = 1e-6
 clip_norm = 1.0
-BPTT_truncate = 28*28
 
 # the data, shuffled and split between train and test sets
 (X_train, y_train), (X_test, y_test) = mnist.load_data()
@@ -55,11 +54,11 @@ Y_test = np_utils.to_categorical(y_test, nb_classes)
 
 print('Evaluate IRNN...')
 model = Sequential()
-model.add(SimpleRNN(input_dim=1, output_dim=hidden_units,
+model.add(SimpleRNN(output_dim=hidden_units,
                     init=lambda shape: normal(shape, scale=0.001),
                     inner_init=lambda shape: identity(shape, scale=1.0),
-                    activation='relu', truncate_gradient=BPTT_truncate))
-model.add(Dense(hidden_units, nb_classes))
+                    activation='relu', input_shape=X_train.shape[1:]))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 rmsprop = RMSprop(lr=learning_rate)
 model.compile(loss='categorical_crossentropy', optimizer=rmsprop)
@@ -73,8 +72,8 @@ print('IRNN test accuracy:', scores[1])
 
 print('Compare to LSTM...')
 model = Sequential()
-model.add(LSTM(1, hidden_units))
-model.add(Dense(hidden_units, nb_classes))
+model.add(LSTM(hidden_units, input_shape=X_train.shape[1:]))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 rmsprop = RMSprop(lr=learning_rate)
 model.compile(loss='categorical_crossentropy', optimizer=rmsprop)

examples/mnist_mlp.py

@@ -12,8 +12,9 @@ from keras.utils import np_utils
 '''
     Train a simple deep NN on the MNIST dataset.
 
-    Get to 98.30% test accuracy after 20 epochs (there is *a lot* of margin for parameter tuning).
-    2 seconds per epoch on a GRID K520 GPU.
+    Get to 98.40% test accuracy after 20 epochs
+    (there is *a lot* of margin for parameter tuning).
+    2 seconds per epoch on a K520 GPU.
 '''
 
 batch_size = 128
@@ -37,19 +38,23 @@ Y_train = np_utils.to_categorical(y_train, nb_classes)
 Y_test = np_utils.to_categorical(y_test, nb_classes)
 
 model = Sequential()
-model.add(Dense(784, 128))
+model.add(Dense(512, input_shape=(784,)))
 model.add(Activation('relu'))
 model.add(Dropout(0.2))
-model.add(Dense(128, 128))
+model.add(Dense(512))
 model.add(Activation('relu'))
 model.add(Dropout(0.2))
-model.add(Dense(128, 10))
+model.add(Dense(10))
 model.add(Activation('softmax'))
 
 rms = RMSprop()
 model.compile(loss='categorical_crossentropy', optimizer=rms)
 
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=2, validation_data=(X_test, Y_test))
-score = model.evaluate(X_test, Y_test, show_accuracy=True, verbose=0)
+model.fit(X_train, Y_train,
+          batch_size=batch_size, nb_epoch=nb_epoch,
+          show_accuracy=True, verbose=2,
+          validation_data=(X_test, Y_test))
+score = model.evaluate(X_test, Y_test,
+                       show_accuracy=True, verbose=0)
 print('Test score:', score[0])
 print('Test accuracy:', score[1])

examples/mnist_transfer_cnn.py

@@ -0,0 +1,122 @@
+from __future__ import absolute_import
+from __future__ import print_function
+import numpy as np
+import datetime
+
+np.random.seed(1337)  # for reproducibility
+
+from keras.datasets import mnist
+from keras.models import Sequential
+from keras.layers.core import Dense, Dropout, Activation, Flatten
+from keras.layers.convolutional import Convolution2D, MaxPooling2D
+from keras.utils import np_utils
+
+'''
+    Transfer learning toy example:
+        1- Train a simple convnet on the MNIST dataset the first 5 digits [0..4].
+        2- Freeze convolutional layers and fine-tune dense layers
+           for the classification of digits [5..9].
+
+    Run on GPU: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python mnist_cnn.py
+
+    Get to 99.8% test accuracy after 5 epochs
+    for the first five digits classifier
+    and 99.2% for the last five digits after transfer + fine-tuning.
+'''
+
+now = datetime.datetime.now
+
+batch_size = 128
+nb_classes = 5
+nb_epoch = 5
+
+# input image dimensions
+img_rows, img_cols = 28, 28
+# number of convolutional filters to use
+nb_filters = 32
+# size of pooling area for max pooling
+nb_pool = 2
+# convolution kernel size
+nb_conv = 3
+
+
+def train_model(model, train, test, nb_classes):
+    X_train = train[0].reshape(train[0].shape[0], 1, img_rows, img_cols)
+    X_test = test[0].reshape(test[0].shape[0], 1, img_rows, img_cols)
+    X_train = X_train.astype("float32")
+    X_test = X_test.astype("float32")
+    X_train /= 255
+    X_test /= 255
+    print('X_train shape:', X_train.shape)
+    print(X_train.shape[0], 'train samples')
+    print(X_test.shape[0], 'test samples')
+
+    # convert class vectors to binary class matrices
+    Y_train = np_utils.to_categorical(train[1], nb_classes)
+    Y_test = np_utils.to_categorical(test[1], nb_classes)
+
+    model.compile(loss='categorical_crossentropy', optimizer='adadelta')
+
+    t = now()
+    model.fit(X_train, Y_train,
+              batch_size=batch_size, nb_epoch=nb_epoch,
+              show_accuracy=True, verbose=1,
+              validation_data=(X_test, Y_test))
+    print('Training time: %s' % (now() - t))
+    score = model.evaluate(X_test, Y_test, show_accuracy=True, verbose=0)
+    print('Test score:', score[0])
+    print('Test accuracy:', score[1])
+
+
+# the data, shuffled and split between train and test sets
+(X_train, y_train), (X_test, y_test) = mnist.load_data()
+
+# create two datasets one with digits below 5 and one with 5 and above
+X_train_lt5 = X_train[y_train < 5]
+y_train_lt5 = y_train[y_train < 5]
+X_test_lt5 = X_test[y_test < 5]
+y_test_lt5 = y_test[y_test < 5]
+
+X_train_gte5 = X_train[y_train >= 5]
+y_train_gte5 = y_train[y_train >= 5] - 5  # make classes start at 0 for
+X_test_gte5 = X_test[y_test >= 5]         # np_utils.to_categorical
+y_test_gte5 = y_test[y_test >= 5] - 5
+
+# define two groups of layers: feature (convolutions) and classification (dense)
+feature_layers = [
+    Convolution2D(nb_filters, nb_conv, nb_conv,
+                  border_mode='valid',
+                  input_shape=(1, img_rows, img_cols)),
+    Activation('relu'),
+    Convolution2D(nb_filters, nb_conv, nb_conv),
+    Activation('relu'),
+    MaxPooling2D(pool_size=(nb_pool, nb_pool)),
+    Dropout(0.25),
+    Flatten(),
+]
+classification_layers = [
+    Dense(128),
+    Activation('relu'),
+    Dropout(0.5),
+    Dense(nb_classes),
+    Activation('softmax')
+]
+
+# create complete model
+model = Sequential()
+for l in feature_layers + classification_layers:
+    model.add(l)
+
+# train model for 5-digit classification [0..4]
+train_model(model,
+            (X_train_lt5, y_train_lt5),
+            (X_test_lt5, y_test_lt5), nb_classes)
+
+# freeze feature layers and rebuild model
+for l in feature_layers:
+    l.trainable = False
+
+# transfer: train dense layers for new classification task [5..9]
+train_model(model,
+            (X_train_gte5, y_train_gte5),
+            (X_test_gte5, y_test_gte5), nb_classes)

examples/neural_turing_machine_copy.py

@@ -0,0 +1,168 @@
+from __future__ import absolute_import
+from __future__ import print_function
+import numpy as np
+np.random.seed(123)
+import matplotlib.pyplot as plt
+
+from theano import function
+
+from keras.models import Sequential
+from keras.layers.core import TimeDistributedDense, Activation
+from keras.layers.recurrent import LSTM
+from keras.optimizers import Adam
+from keras.utils import generic_utils
+
+from keras.layers.ntm import NeuralTuringMachine as NTM
+
+"""
+Copy Problem defined in Graves et. al [0]
+
+Training data is made of sequences with length 1 to 20. 
+Test data are sequences of length 100.
+The model is tested every 500 weight updates.
+After about 3500 updates, the accuracy jumps from around 50% to >90%.
+
+Estimated compile time: 12 min
+Estimated time to train Neural Turing Machine and 3 layer LSTM on an NVidia GTX 680: 2h
+
+[0]: http://arxiv.org/pdf/1410.5401v2.pdf
+"""
+
+batch_size = 100
+
+h_dim = 128
+n_slots = 128
+m_length = 20
+input_dim = 8
+lr = 1e-3
+clipvalue = 10
+
+##### Neural Turing Machine ######
+
+ntm = NTM(h_dim, n_slots=n_slots, m_length=m_length, shift_range=3,
+          inner_rnn='lstm', return_sequences=True, input_dim=input_dim)
+model = Sequential()
+model.add(ntm)
+model.add(TimeDistributedDense(input_dim))
+model.add(Activation('sigmoid'))
+
+sgd = Adam(lr=lr, clipvalue=clipvalue)
+model.compile(loss='binary_crossentropy', optimizer=sgd)
+
+# LSTM - Run this for comparison
+
+sgd2 = Adam(lr=lr, clipvalue=clipvalue)
+lstm = Sequential()
+lstm.add(LSTM(input_dim=input_dim, output_dim=h_dim*2, return_sequences=True))
+lstm.add(LSTM(output_dim=h_dim*2, return_sequences=True))
+lstm.add(LSTM(output_dim=h_dim*2, return_sequences=True))
+lstm.add(TimeDistributedDense(input_dim))
+lstm.add(Activation('sigmoid'))
+
+lstm.compile(loss='binary_crossentropy', optimizer=sgd)
+
+###### DATASET ########
+
+def get_sample(batch_size=128, n_bits=8, max_size=20, min_size=1):
+    # generate samples with random length
+    inp = np.zeros((batch_size, 2*max_size-1, n_bits))
+    out = np.zeros((batch_size, 2*max_size-1, n_bits))
+    sw = np.zeros((batch_size, 2*max_size-1, 1))
+    for i in range(batch_size):
+        t = np.random.randint(low=min_size, high=max_size)
+        x = np.random.uniform(size=(t, n_bits)) > .5
+        for j,f in enumerate(x.sum(axis=-1)): # remove fake flags
+            if f>=n_bits:
+                x[j, :] = 0.
+        del_flag = np.ones((1, n_bits))
+        inp[i, :t+1] = np.concatenate([x, del_flag], axis=0)
+        out[i, t+1:(2*t+1)] = x
+        sw[i, t+1:(2*t+1)] = 1
+    return inp, out, sw
+
+def show_pattern(inp, out, sw, file_name='ntm_output.png'):
+    ''' Helper function to visualize results '''
+    plt.figure(figsize=(10, 10))
+    plt.subplot(131)
+    plt.imshow(inp>.5)
+    plt.subplot(132)
+    plt.imshow(out>.5)
+    plt.subplot(133)
+    plt.imshow(sw>.5)
+    plt.savefig(file_name)
+    plt.close()
+
+# Show data example:
+inp, out, sw = get_sample(1, 8, 20)
+
+plt.subplot(131)
+plt.title('input')
+plt.imshow(inp[0], cmap='gray')
+plt.subplot(132)
+plt.title('desired')
+plt.imshow(out[0], cmap='gray')
+plt.subplot(133)
+plt.title('sample_weight')
+plt.imshow(sw[0], cmap='gray')
+
+# training uses sequences of length 1 to 20. Test uses series of length 100.
+def test_model(model, file_name, min_size=100):
+    I, V, sw = get_sample(batch_size=500, n_bits=input_dim, max_size=min_size+1, min_size=min_size)
+    Y = np.asarray(model.predict(I, batch_size=100) > .5).astype('float64')
+    acc = (V[:, -min_size:, :] == Y[:, -min_size:, :]).mean() * 100
+    show_pattern(Y[0], V[0], sw[0], file_name)
+    return acc
+
+##### TRAIN ######
+nb_epoch = 4000
+progbar = generic_utils.Progbar(nb_epoch)
+for e in range(nb_epoch):
+    I, V, sw = get_sample(n_bits=input_dim, max_size=20, min_size=1, batch_size=100)
+
+    loss1 = model.train_on_batch(I, V, sample_weight=sw)
+    loss2 = lstm.train_on_batch(I, V, sample_weight=sw)
+
+    progbar.add(1, values=[("NTM", loss1), ("LSTM", loss2)])
+
+    if e % 500 == 0:
+        print("")
+        acc1 = test_model(model, 'ntm.png')
+        acc2 = test_model(lstm, 'lstm.png')
+        print("NTM  test acc: {}".format(acc1))
+        print("LSTM test acc: {}".format(acc2))
+
+##### VISUALIZATION #####
+X = model.get_input()
+Y = ntm.get_full_output()[0:3]  # (memory over time, read_vectors, write_vectors)
+F = function([X], Y, allow_input_downcast=True)
+
+inp, out, sw = get_sample(1, 8, 21, 20)
+mem, read, write = F(inp.astype('float32'))
+Y = model.predict(inp)
+
+plt.figure(figsize=(15, 12))
+
+plt.subplot(221)
+plt.imshow(write[0])
+plt.xlabel('memory location')
+plt.ylabel('time')
+plt.title('write')
+
+plt.subplot(222)
+plt.imshow(read[0])
+plt.title('read')
+
+plt.subplot(223)
+plt.title('desired')
+plt.imshow(out[0])
+
+plt.subplot(224)
+plt.imshow(Y[0]>.5)
+plt.title('output')
+
+plt.figure(figsize=(15, 10))
+plt.subplot(325)
+plt.ylabel('time')
+plt.xlabel('location')
+plt.title('memory evolving in time (avg value per location)')
+plt.imshow(mem[0].mean(axis=-1))

examples/reuters_mlp.py

@@ -45,10 +45,10 @@ print('Y_test shape:', Y_test.shape)
 
 print("Building model...")
 model = Sequential()
-model.add(Dense(max_words, 512))
+model.add(Dense(512, input_shape=(max_words,)))
 model.add(Activation('relu'))
 model.add(Dropout(0.5))
-model.add(Dense(512, nb_classes))
+model.add(Dense(nb_classes))
 model.add(Activation('softmax'))
 
 model.compile(loss='categorical_crossentropy', optimizer='adam')

examples/skipgram_word_embeddings.py

@@ -1,222 +0,0 @@
-
-'''
-    We loop over words in a dataset, and for each word, we look at a context window around the word.
-    We generate pairs of (pivot_word, other_word_from_same_context) with label 1,
-    and pairs of (pivot_word, random_word) with label 0 (skip-gram method).
-
-    We use the layer WordContextProduct to learn embeddings for the word couples,
-    and compute a proximity score between the embeddings (= p(context|word)),
-    trained with our positive and negative labels.
-
-    We then use the weights computed by WordContextProduct to encode words
-    and demonstrate that the geometry of the embedding space
-    captures certain useful semantic properties.
-
-    Read more about skip-gram in this particularly gnomic paper by Mikolov et al.:
-        http://arxiv.org/pdf/1301.3781v3.pdf
-
-    Note: you should run this on GPU, otherwise training will be quite slow.
-    On a EC2 GPU instance, expect 3 hours per 10e6 comments (~10e8 words) per epoch with dim_proj=256.
-    Should be much faster on a modern GPU.
-
-    GPU command:
-        THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python skipgram_word_embeddings.py
-
-    Dataset: 5,845,908 Hacker News comments.
-    Obtain the dataset at:
-        https://mega.co.nz/#F!YohlwD7R!wec0yNO86SeaNGIYQBOR0A
-        (HNCommentsAll.1perline.json.bz2)
-'''
-from __future__ import absolute_import
-from __future__ import print_function
-
-import numpy as np
-import theano
-import six.moves.cPickle
-import os, re, json
-
-from keras.preprocessing import sequence, text
-from keras.optimizers import SGD, RMSprop, Adagrad
-from keras.utils import np_utils, generic_utils
-from keras.models import Sequential
-from keras.layers.embeddings import WordContextProduct, Embedding
-from six.moves import range
-from six.moves import zip
-
-max_features = 50000  # vocabulary size: top 50,000 most common words in data
-skip_top = 100  # ignore top 100 most common words
-nb_epoch = 1
-dim_proj = 256  # embedding space dimension
-
-save = True
-load_model = False
-load_tokenizer = False
-train_model = True
-save_dir = os.path.expanduser("~/.keras/models")
-if not os.path.exists(save_dir):
-    os.makedirs(save_dir)
-model_load_fname = "HN_skipgram_model.pkl"
-model_save_fname = "HN_skipgram_model.pkl"
-tokenizer_fname = "HN_tokenizer.pkl"
-
-data_path = os.path.expanduser("~/")+"HNCommentsAll.1perline.json"
-
-# text preprocessing utils
-html_tags = re.compile(r'<.*?>')
-to_replace = [('&#x27;', "'")]
-hex_tags = re.compile(r'&.*?;')
-
-
-def clean_comment(comment):
-    c = str(comment.encode("utf-8"))
-    c = html_tags.sub(' ', c)
-    for tag, char in to_replace:
-        c = c.replace(tag, char)
-    c = hex_tags.sub(' ', c)
-    return c
-
-
-def text_generator(path=data_path):
-    f = open(path)
-    for i, l in enumerate(f):
-        comment_data = json.loads(l)
-        comment_text = comment_data["comment_text"]
-        comment_text = clean_comment(comment_text)
-        if i % 10000 == 0:
-            print(i)
-        yield comment_text
-    f.close()
-
-# model management
-if load_tokenizer:
-    print('Load tokenizer...')
-    tokenizer = six.moves.cPickle.load(open(os.path.join(save_dir, tokenizer_fname), 'rb'))
-else:
-    print("Fit tokenizer...")
-    tokenizer = text.Tokenizer(nb_words=max_features)
-    tokenizer.fit_on_texts(text_generator())
-    if save:
-        print("Save tokenizer...")
-        if not os.path.exists(save_dir):
-            os.makedirs(save_dir)
-        six.moves.cPickle.dump(tokenizer, open(os.path.join(save_dir, tokenizer_fname), "wb"))
-
-# training process
-if train_model:
-    if load_model:
-        print('Load model...')
-        model = six.moves.cPickle.load(open(os.path.join(save_dir, model_load_fname), 'rb'))
-    else:
-        print('Build model...')
-        model = Sequential()
-        model.add(WordContextProduct(max_features, proj_dim=dim_proj, init="uniform"))
-        model.compile(loss='mse', optimizer='rmsprop')
-
-    sampling_table = sequence.make_sampling_table(max_features)
-
-    for e in range(nb_epoch):
-        print('-'*40)
-        print('Epoch', e)
-        print('-'*40)
-
-        progbar = generic_utils.Progbar(tokenizer.document_count)
-        samples_seen = 0
-        losses = []
-
-        for i, seq in enumerate(tokenizer.texts_to_sequences_generator(text_generator())):
-            # get skipgram couples for one text in the dataset
-            couples, labels = sequence.skipgrams(seq, max_features, window_size=4, negative_samples=1., sampling_table=sampling_table)
-            if couples:
-                # one gradient update per sentence (one sentence = a few 1000s of word couples)
-                X = np.array(couples, dtype="int32")
-                loss = model.train(X, labels)
-                losses.append(loss)
-                if len(losses) % 100 == 0:
-                    progbar.update(i, values=[("loss", np.mean(losses))])
-                    losses = []
-                samples_seen += len(labels)
-        print('Samples seen:', samples_seen)
-    print("Training completed!")
-
-    if save:
-        print("Saving model...")
-        if not os.path.exists(save_dir):
-            os.makedirs(save_dir)
-        six.moves.cPickle.dump(model, open(os.path.join(save_dir, model_save_fname), "wb"))
-
-
-print("It's test time!")
-
-# recover the embedding weights trained with skipgram:
-weights = model.layers[0].get_weights()[0]
-
-# we no longer need this
-del model
-
-weights[:skip_top] = np.zeros((skip_top, dim_proj))
-norm_weights = np_utils.normalize(weights)
-
-word_index = tokenizer.word_index
-reverse_word_index = dict([(v, k) for k, v in list(word_index.items())])
-word_index = tokenizer.word_index
-
-
-def embed_word(w):
-    i = word_index.get(w)
-    if (not i) or (i < skip_top) or (i >= max_features):
-        return None
-    return norm_weights[i]
-
-
-def closest_to_point(point, nb_closest=10):
-    proximities = np.dot(norm_weights, point)
-    tups = list(zip(list(range(len(proximities))), proximities))
-    tups.sort(key=lambda x: x[1], reverse=True)
-    return [(reverse_word_index.get(t[0]), t[1]) for t in tups[:nb_closest]]
-
-
-def closest_to_word(w, nb_closest=10):
-    i = word_index.get(w)
-    if (not i) or (i < skip_top) or (i >= max_features):
-        return []
-    return closest_to_point(norm_weights[i].T, nb_closest)
-
-
-''' the resuls in comments below were for:
-    5.8M HN comments
-    dim_proj = 256
-    nb_epoch = 2
-    optimizer = rmsprop
-    loss = mse
-    max_features = 50000
-    skip_top = 100
-    negative_samples = 1.
-    window_size = 4
-    and frequency subsampling of factor 10e-5.
-'''
-
-words = [
-    "article",  # post, story, hn, read, comments
-    "3",  # 6, 4, 5, 2
-    "two",  # three, few, several, each
-    "great",  # love, nice, working, looking
-    "data",  # information, memory, database
-    "money",  # company, pay, customers, spend
-    "years",  # ago, year, months, hours, week, days
-    "android",  # ios, release, os, mobile, beta
-    "javascript",  # js, css, compiler, library, jquery, ruby
-    "look",  # looks, looking
-    "business",  # industry, professional, customers
-    "company",  # companies, startup, founders, startups
-    "after",  # before, once, until
-    "own",  # personal, our, having
-    "us",  # united, country, american, tech, diversity, usa, china, sv
-    "using",  # javascript, js, tools (lol)
-    "here",  # hn, post, comments
-]
-
-for w in words:
-    res = closest_to_word(w)
-    print('====', w)
-    for r in res:
-        print(r)

keras/__init__.py

@@ -0,0 +1,15 @@
+
+"""
+Keras: Theano-based Deep Learning library
+==================================
+Keras is a minimalist, highly modular neural network library in 
+the spirit of Torch, written in Python / Theano so as not to have 
+to deal with the dearth of ecosystem in Lua. It was developed with 
+a focus on enabling fast experimentation. Being able to go from 
+idea to result with the least possible delay is key to doing 
+good research.
+
+See http://keras.io/
+"""
+
+__version__ = '0.2.0'

keras/activations.py

@@ -1,35 +1,40 @@
 from __future__ import absolute_import
-import theano.tensor as T
+from . import backend as K
 
 
 def softmax(x):
-    return T.nnet.softmax(x.reshape((-1, x.shape[-1]))).reshape(x.shape)
-
-
-def time_distributed_softmax(x):
-    import warnings
-    warnings.warn("time_distributed_softmax is deprecated. Just use softmax!", DeprecationWarning)
-    return softmax(x)
+    ndim = K.ndim(x)
+    if ndim == 2:
+        return K.softmax(x)
+    elif ndim == 3:
+        # apply softmax to each timestep
+        def step(x, states):
+            return K.softmax(x), []
+        last_output, outputs, states = K.rnn(step, x, [], masking=False)
+        return outputs
+    else:
+        raise Exception('Cannot apply softmax to a tensor that is not 2D or 3D. ' +
+                        'Here, ndim=' + str(ndim))
 
 
 def softplus(x):
-    return T.nnet.softplus(x)
+    return K.softplus(x)
 
 
-def relu(x):
-    return (x + abs(x)) / 2.0
+def relu(x, alpha=0., max_value=None):
+    return K.relu(x, alpha=alpha, max_value=max_value)
 
 
 def tanh(x):
-    return T.tanh(x)
+    return K.tanh(x)
 
 
 def sigmoid(x):
-    return T.nnet.sigmoid(x)
+    return K.sigmoid(x)
 
 
 def hard_sigmoid(x):
-    return T.nnet.hard_sigmoid(x)
+    return K.hard_sigmoid(x)
 
 
 def linear(x):

keras/backend/__init__.py

@@ -0,0 +1,39 @@
+from __future__ import absolute_import
+from __future__ import print_function
+import os
+import json
+from .common import epsilon, floatx, set_epsilon, set_floatx
+
+_keras_dir = os.path.expanduser(os.path.join('~', '.keras'))
+if not os.path.exists(_keras_dir):
+    os.makedirs(_keras_dir)
+
+_BACKEND = 'theano'
+_config_path = os.path.expanduser(os.path.join('~', '.keras', 'keras.json'))
+if os.path.exists(_config_path):
+    _config = json.load(open(_config_path))
+    _floatx = _config.get('floatx', floatx())
+    assert _floatx in {'float32', 'float64'}
+    _epsilon = _config.get('epsilon', epsilon())
+    assert type(_epsilon) == float
+    _backend = _config.get('backend', _BACKEND)
+    assert _backend in {'theano', 'tensorflow'}
+
+    set_floatx(_floatx)
+    set_epsilon(_epsilon)
+    _BACKEND = _backend
+else:
+    # save config file, for easy edition
+    _config = {'floatx': floatx(),
+               'epsilon': epsilon(),
+               'backend': _BACKEND}
+    json.dump(_config, open(_config_path, 'w'))
+
+if _BACKEND == 'theano':
+    print('Using Theano backend.')
+    from .theano_backend import *
+elif _BACKEND == 'tensorflow':
+    print('Using TensorFlow backend.')
+    from .tensorflow_backend import *
+else:
+    raise Exception('Unknown backend: ' + str(backend))

keras/backend/common.py

@@ -0,0 +1,31 @@
+import numpy as np
+
+# the type of float to use throughout the session.
+_FLOATX = 'float32'
+_EPSILON = 10e-8
+
+
+def epsilon():
+    return _EPSILON
+
+
+def set_epsilon(e):
+    global _EPSILON
+    _EPSILON = e
+
+
+def floatx():
+    return _FLOATX
+
+
+def set_floatx(floatx):
+    global _FLOATX
+    if floatx not in {'float32', 'float64'}:
+        raise Exception('Unknown floatx type: ' + str(floatx))
+    _FLOATX = floatx
+
+
+def cast_to_floatx(x):
+    '''Cast a Numpy array to floatx.
+    '''
+    return np.asarray(x, dtype=_FLOATX)

keras/backend/tensorflow_backend.py

@@ -0,0 +1,602 @@
+import tensorflow as tf
+import numpy as np
+from .common import _FLOATX, _EPSILON
+
+# INTERNAL UTILS
+
+_SESSION = None
+
+
+def _get_session():
+    global _SESSION
+    if _SESSION is None:
+        _SESSION = tf.Session('')
+    return _SESSION
+
+
+def _set_session(session):
+    global _SESSION
+    _SESSION = session
+
+
+# VARIABLE MANIPULATION
+
+def variable(value, dtype=_FLOATX, name=None):
+    v = tf.Variable(np.asarray(value, dtype=dtype), name=name)
+    _get_session().run(v.initializer)
+    return v
+
+
+def placeholder(shape=None, ndim=None, dtype=_FLOATX, name=None):
+    if not shape:
+        if ndim:
+            shape = [None for _ in range(ndim)]
+    return tf.placeholder(dtype, shape=shape, name=name)
+
+
+def shape(x):
+    return x.get_shape()
+
+
+def ndim(x):
+    return len(x.get_shape())
+
+
+def eval(x):
+    '''Run a graph.
+    '''
+    return x.eval(session=_get_session())
+
+
+def zeros(shape, dtype=_FLOATX, name=None):
+    return variable(np.zeros(shape), dtype, name)
+
+
+def ones(shape, dtype=_FLOATX, name=None):
+    return variable(np.ones(shape), dtype, name)
+
+
+def ones_like(x, name=None):
+    return tf.ones_like(x)
+
+
+def zeros_like(x, name=None):
+    return tf.zeros_like(x)
+
+
+def count_params(x):
+    '''Return number of scalars in a tensor.
+    '''
+    shape = x.get_shape()
+    return np.prod([shape[i]._value for i in range(len(shape))])
+
+
+def cast(x, dtype):
+    return tf.cast(x, dtype)
+
+
+# LINEAR ALGEBRA
+
+def dot(x, y):
+    return tf.matmul(x, y)
+
+
+def transpose(x):
+    return tf.transpose(x)
+
+
+def gather(reference, indices):
+    '''reference: a tensor.
+    indices: an int tensor of indices.
+
+    Return: a tensor of same type as reference.
+    '''
+    return tf.gather(reference, indices)
+
+
+# ELEMENT-WISE OPERATIONS
+
+def max(x, axis=None, keepdims=False):
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    return tf.reduce_max(x, reduction_indices=axis, keep_dims=keepdims)
+
+
+def min(x, axis=None, keepdims=False):
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    return tf.reduce_min(x, reduction_indices=axis, keep_dims=keepdims)
+
+
+def sum(x, axis=None, keepdims=False):
+    '''Sum of the values in a tensor, alongside the specified axis.
+    '''
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    return tf.reduce_sum(x, reduction_indices=axis, keep_dims=keepdims)
+
+
+def prod(x, axis=None, keepdims=False):
+    '''Multiply the values in a tensor, alongside the specified axis.
+    '''
+    return tf.reduce_prod(x, reduction_indices=axis, keep_dims=keepdims)
+
+
+def std(x, axis=None, keepdims=False):
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    if x.dtype.base_dtype == tf.bool:
+        x = tf.cast(x, _FLOATX)
+    m = tf.reduce_mean(x, reduction_indices=axis, keep_dims=keepdims)
+    devs_squared = tf.square(x - m)
+    return tf.sqrt(tf.reduce_mean(devs_squared,
+                                  reduction_indices=axis,
+                                  keep_dims=keepdims))
+
+
+def mean(x, axis=None, keepdims=False):
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    if x.dtype.base_dtype == tf.bool:
+        x = tf.cast(x, _FLOATX)
+    return tf.reduce_mean(x, reduction_indices=axis, keep_dims=keepdims)
+
+
+def any(x, axis=None, keepdims=False):
+    '''Bitwise reduction (logical OR).
+
+    Return array of int8 (0s and 1s).
+    '''
+    if axis is not None and axis < 0:
+        axis = axis % len(x.get_shape())
+    x = tf.cast(x, tf.bool)
+    x = tf.reduce_any(x, reduction_indices=axis, keep_dims=keepdims)
+    return tf.cast(x, tf.int8)
+
+
+def argmax(x, axis=-1):
+    if axis < 0:
+        axis = axis % len(x.get_shape())
+    return tf.argmax(x, axis)
+
+
+def argmin(x, axis=-1):
+    if axis < 0:
+        axis = axis % len(x.get_shape())
+    return tf.argmin(x, axis)
+
+
+def square(x):
+    return tf.square(x)
+
+
+def abs(x):
+    return tf.abs(x)
+
+
+def sqrt(x):
+    x = tf.clip_by_value(x, tf.cast(0., dtype=_FLOATX),
+                         tf.cast(np.inf, dtype=_FLOATX))
+    return tf.sqrt(x)
+
+
+def exp(x):
+    return tf.exp(x)
+
+
+def log(x):
+    return tf.log(x)
+
+
+def round(x):
+    return tf.round(x)
+
+
+def pow(x, a):
+    return tf.pow(x, a)
+
+
+def clip(x, min_value, max_value):
+    if max_value < min_value:
+        max_value = min_value
+    return tf.clip_by_value(x, tf.cast(min_value, dtype=_FLOATX),
+                            tf.cast(max_value, dtype=_FLOATX))
+
+
+def equal(x, y):
+    return tf.equal(x, y)
+
+
+def maximum(x, y):
+    return tf.maximum(x, y)
+
+
+def minimum(x, y):
+    return tf.minimum(x, y)
+
+
+# SHAPE OPERATIONS
+
+def concatenate(tensors, axis=-1):
+    if axis < 0:
+        axis = axis % len(tensors[0].get_shape())
+    return tf.concat(axis, tensors)
+
+
+def reshape(x, shape):
+    return tf.reshape(x, shape)
+
+
+def permute_dimensions(x, pattern):
+    '''Transpose dimensions.
+
+    pattern should be a tuple or list of
+    dimension indices, e.g. [0, 2, 1].
+    '''
+    return tf.transpose(x, perm=pattern)
+
+
+def repeat(x, n):
+    '''Repeat a 2D tensor:
+
+    if x has shape (samples, dim) and n=2,
+    the output will have shape (samples, 2, dim)
+    '''
+    tensors = [x] * n
+    stacked = tf.pack(tensors)
+    return tf.transpose(stacked, (1, 0, 2))
+
+
+def tile(x, n):
+    return tf.tile(x, n)
+
+
+def flatten(x):
+    '''Turn a n-D tensor into a 2D tensor where
+    the first dimension is conserved.
+    '''
+    x = tf.reshape(x, [-1, np.prod(x.get_shape()[1:].as_list())])
+    return x
+
+
+def expand_dims(x, dim=-1):
+    '''Add a 1-sized dimension at index "dim".
+    '''
+    return tf.expand_dims(x, dim)
+
+
+def squeeze(x, axis):
+    '''Remove a 1-dimension from the tensor at index "axis".
+    '''
+    return tf.squeeze(x, [axis])
+
+
+def temporal_padding(x, padding=1):
+    '''Pad the middle dimension of a 3D tensor
+    with "padding" zeros left and right.
+
+    Appologies for the inane API, but Theano makes this
+    really hard.
+    '''
+    pattern = [[0, 0], [padding, padding], [0, 0]]
+    return tf.pad(x, pattern)
+
+
+def spatial_2d_padding(x, padding=(1, 1), dim_ordering='th'):
+    '''Pad the 2nd and 3rd dimensions of a 4D tensor
+    with "padding[0]" and "padding[1]" (resp.) zeros left and right.
+    '''
+    if dim_ordering == 'th':
+        pattern = [[0, 0], [0, 0],
+                   [padding[0], padding[0]], [padding[1], padding[1]]]
+    else:
+        pattern = [[0, 0],
+                   [padding[0], padding[0]], [padding[1], padding[1]],
+                   [0, 0]]
+    return tf.pad(x, pattern)
+
+
+# VALUE MANIPULATION
+
+def get_value(x):
+    '''Technically the same as eval() for TF.
+    '''
+    return x.eval(session=_get_session())
+
+
+def set_value(x, value):
+    tf.assign(x, np.asarray(value)).op.run(session=_get_session())
+
+
+# GRAPH MANIPULATION
+
+class Function(object):
+
+    def __init__(self, inputs, outputs, updates=[]):
+        self.inputs = list(inputs)
+        self.outputs = list(outputs)
+        with tf.control_dependencies(self.outputs):
+            self.updates = [tf.assign(p, new_p) for (p, new_p) in updates]
+
+    def __call__(self, inputs):
+        names = [v.name for v in self.inputs]
+        feed_dict = dict(zip(names, inputs))
+        session = _get_session()
+        updated = session.run(self.outputs + self.updates, feed_dict=feed_dict)
+        return updated[:len(self.outputs)]
+
+
+def function(inputs, outputs, updates=[]):
+    return Function(inputs, outputs, updates=updates)
+
+
+def gradients(loss, variables):
+    return tf.gradients(loss, variables)
+
+
+# CONTROL FLOW
+
+def rnn(step_function, inputs, initial_states,
+        go_backwards=False, masking=True):
+    '''Iterates over the time dimension of a tensor.
+
+    Parameters
+    ----------
+    inputs: tensor of temporal data of shape (samples, time, ...)
+        (at least 3D).
+    step_function:
+        Parameters:
+            input: tensor with shape (samples, ...) (no time dimension),
+                representing input for the batch of samples at a certain
+                time step.
+            states: list of tensors.
+        Returns:
+            output: tensor with shape (samples, ...) (no time dimension),
+            new_states: list of tensors, same length and shapes
+                as 'states'.
+    initial_states: tensor with shape (samples, ...) (no time dimension),
+        containing the initial values for the states used in
+        the step function.
+    go_backwards: boolean. If True, do the iteration over
+        the time dimension in reverse order.
+    masking: boolean. If true, any input timestep inputs[s, i]
+        that is all-zeros will be skipped (states will be passed to
+        the next step unchanged) and the corresponding output will
+        be all zeros.
+
+    Returns
+    -------
+    A tuple (last_output, outputs, new_states).
+        last_output: the latest output of the rnn, of shape (samples, ...)
+        outputs: tensor with shape (samples, time, ...) where each
+            entry outputs[s, t] is the output of the step function
+            at time t for sample s.
+        new_states: list of tensors, latest states returned by
+            the step function, of shape (samples, ...).
+    '''
+    inputs = tf.transpose(inputs, (1, 0, 2))
+    input_list = tf.unpack(inputs)
+
+    states = initial_states
+    successive_states = []
+    successive_outputs = []
+    if go_backwards:
+        input_list.reverse()
+    for input in input_list:
+        output, new_states = step_function(input, states)
+        if masking:
+            # for now we raise an exception because tf.reduce_any will not work
+            raise Exception("Masking is Theano-only for the time being.")
+
+            # if all-zero input timestep, return
+            # all-zero output and unchanged states
+            switch = tf.reduce_any(input)
+            output = tf.control_flow_ops.cond(switch,
+                                              lambda: output,
+                                              lambda: 0. * output)
+            return_states = []
+            for state, new_state in zip(states, new_states):
+                return_states.append(tf.control_flow_ops.cond(switch,
+                                                              lambda: new_state,
+                                                              lambda: state))
+            states = return_states
+        else:
+            states = new_states
+        successive_outputs.append(output)
+        successive_states.append(states)
+
+    last_output = successive_outputs[-1]
+    outputs = tf.pack(successive_outputs)
+    new_states = successive_states[-1]
+
+    outputs = tf.transpose(outputs, (1, 0, 2))
+    return last_output, outputs, states
+
+
+def switch(condition, then_expression, else_expression):
+    '''condition: scalar tensor.
+    '''
+    return tf.control_flow_ops.cond(condition,
+                                    lambda: then_expression,
+                                    lambda: else_expression)
+
+
+# NN OPERATIONS
+
+def relu(x, alpha=0., max_value=None):
+    '''ReLU.
+
+    alpha: slope of negative section.
+    '''
+    negative_part = tf.nn.relu(-x)
+    x = tf.nn.relu(x)
+    if max_value is not None:
+        x = tf.clip_by_value(x, tf.cast(0., dtype=_FLOATX),
+                             tf.cast(max_value, dtype=_FLOATX))
+    x -= tf.constant(alpha, dtype=_FLOATX) * negative_part
+    return x
+
+
+def softmax(x):
+    return tf.nn.softmax(x)
+
+
+def softplus(x):
+    return tf.nn.softplus(x)
+
+
+def categorical_crossentropy(output, target, from_logits=False):
+    '''Note: tf.nn.softmax_cross_entropy_with_logits
+    expects logits, Keras expects probabilities.
+    '''
+    if not from_logits:
+        # scale preds so that the class probas of each sample sum to 1
+        output /= tf.reduce_sum(output,
+                                reduction_indices=len(output.get_shape())-1,
+                                keep_dims=True)
+        # manual computation of crossentropy
+        output = tf.clip_by_value(output, tf.cast(_EPSILON, dtype=_FLOATX),
+                                  tf.cast(1.-_EPSILON, dtype=_FLOATX))
+        return - tf.reduce_sum(target * tf.log(output),
+                               reduction_indices=len(output.get_shape())-1)
+    else:
+        return tf.nn.softmax_cross_entropy_with_logits(output, target)
+
+
+def binary_crossentropy(output, target, from_logits=False):
+    '''Note: tf.nn.sigmoid_cross_entropy_with_logits
+    expects logits, Keras expects probabilities.
+    '''
+    if not from_logits:
+        # transform back to logits
+        output = tf.clip_by_value(output, tf.cast(_EPSILON, dtype=_FLOATX),
+                                  tf.cast(1.-_EPSILON, dtype=_FLOATX))
+        output = tf.log(output / (1 - output))
+    return tf.nn.sigmoid_cross_entropy_with_logits(output, target)
+
+
+def sigmoid(x):
+    return tf.nn.sigmoid(x)
+
+
+def hard_sigmoid(x):
+    x = (0.2 * x) + 0.5
+    x = tf.clip_by_value(x, tf.cast(0., dtype=_FLOATX),
+                         tf.cast(1., dtype=_FLOATX))
+    return x
+
+
+def tanh(x):
+    return tf.nn.tanh(x)
+
+
+def dropout(x, level, seed=None):
+    retain_prob = 1. - level
+    if seed is None:
+        seed = np.random.randint(10e6)
+    # the dummy 1. works around a TF bug
+    # (float32_ref vs. float32 incomptability)
+    return tf.nn.dropout(x * 1., retain_prob, seed=seed)
+
+
+# CONVOLUTIONS
+
+
+def conv2d(x, kernel, strides=(1, 1), border_mode='valid', dim_ordering='th'):
+    '''
+    Run on cuDNN if available.
+    border_mode: string, "same" or "valid".
+    dim_ordering: whether to use Theano or TensorFlow dimension ordering
+    in inputs/kernels/ouputs.
+    '''
+    if border_mode == 'same':
+        padding = 'SAME'
+    elif border_mode == 'valid':
+        padding = 'VALID'
+    else:
+        raise Exception('Invalid border mode: ' + str(border_mode))
+
+    strides = (1,) + strides + (1,)
+
+    if _FLOATX == 'float64':
+        # tf conv2d only supports float32
+        x = tf.cast(x, 'float32')
+        kernel = tf.cast(kernel, 'float32')
+
+    if dim_ordering == 'th':
+        # TF uses the last dimension as channel dimension,
+        # instead of the 2nd one.
+        # TH input shape: (samples, input_depth, rows, cols)
+        # TF input shape: (samples, rows, cols, input_depth)
+        # TH kernel shape: (depth, input_depth, rows, cols)
+        # TF kernel shape: (rows, cols, input_depth, depth)
+        x = tf.transpose(x, (0, 2, 3, 1))
+        kernel = tf.transpose(kernel, (2, 3, 1, 0))
+        x = tf.nn.conv2d(x, kernel, strides, padding=padding)
+        x = tf.transpose(x, (0, 3, 1, 2))
+    elif dim_ordering == 'tf':
+        x = tf.nn.conv2d(x, kernel, strides, padding=padding)
+    else:
+        raise Exception('Unknown dim_ordering: ' + str(dim_ordering))
+
+    if _FLOATX == 'float64':
+        x = tf.cast(x, 'float64')
+    return x
+
+
+def maxpool2d(x, pool_size, strides=(1, 1),
+              border_mode='valid', dim_ordering='th'):
+    '''
+    pool_size: tuple of 2 integers.
+    strides: tuple of 2 integers.
+    border_mode: one of "valid", "same".
+    dim_ordering: one of "th", "tf".
+    '''
+    if border_mode == 'same':
+        padding = 'SAME'
+    elif border_mode == 'valid':
+        padding = 'VALID'
+    else:
+        raise Exception('Invalid border mode: ' + str(border_mode))
+
+    strides = (1,) + strides + (1,)
+    pool_size = (1,) + pool_size + (1,)
+
+    if _FLOATX == 'float64':
+        # tf max_pool only supports float32
+        x = tf.cast(x, 'float32')
+
+    if dim_ordering == 'th':
+        # TF uses the last dimension as channel dimension,
+        # instead of the 2nd one.
+        # TH input shape: (samples, input_depth, rows, cols)
+        # TF input shape: (samples, rows, cols, input_depth)
+        # TH kernel shape: (depth, input_depth, rows, cols)
+        # TF kernel shape: (rows, cols, input_depth, depth)
+        x = tf.transpose(x, (0, 2, 3, 1))
+        x = tf.nn.max_pool(x, pool_size, strides, padding=padding)
+        x = tf.transpose(x, (0, 3, 1, 2))
+    elif dim_ordering == 'tf':
+        x = tf.nn.max_pool(x, pool_size, strides, padding=padding)
+    else:
+        raise Exception('Unknown dim_ordering: ' + str(dim_ordering))
+
+    if _FLOATX == 'float64':
+        x = tf.cast(x, 'float64')
+    return x
+
+
+# RANDOMNESS
+
+def random_normal(shape, mean=0.0, std=1.0, dtype=_FLOATX, seed=None):
+    if seed is None:
+        seed = np.random.randint(10e6)
+    return tf.random_normal(shape, mean=mean, stddev=std,
+                            dtype=dtype, seed=seed)
+
+
+def random_uniform(shape, low=0.0, high=1.0, dtype=_FLOATX, seed=None):
+    if seed is None:
+        seed = np.random.randint(10e6)
+    return tf.random_uniform(shape, minval=low, maxval=high,
+                             dtype=dtype, seed=seed)

keras/backend/theano_backend.py

@@ -0,0 +1,622 @@
+import theano
+from theano import tensor as T
+from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams
+from theano.tensor.signal import downsample
+import numpy as np
+from .common import _FLOATX, _EPSILON
+
+
+# INTERNAL UTILS
+theano.config.floatX = _FLOATX
+
+
+def _on_gpu():
+    '''Returns whether the session is set to
+    run on GPU or not (i.e. on CPU).
+    '''
+    return theano.config.device[:3] == 'gpu'
+
+
+if _on_gpu():
+    '''Import cuDNN only if running on GPU:
+    not having Cuda install should not
+    prevent from running the present code.
+    '''
+    from theano.sandbox.cuda import dnn
+
+
+# VARIABLE MANIPULATION
+
+def variable(value, dtype=_FLOATX, name=None):
+    '''Instantiate a tensor variable.
+    '''
+    value = np.asarray(value, dtype=dtype)
+    return theano.shared(value=value, name=name, strict=False)
+
+
+def placeholder(shape=None, ndim=None, dtype=_FLOATX, name=None):
+    '''Instantiate an input data placeholder variable.
+    '''
+    if shape is None and ndim is None:
+        raise Exception('Specify either a shape or ndim value.')
+    if shape is not None:
+        ndim = len(shape)
+    if ndim == 0:
+        return T.scalar(name=name, dtype=dtype)
+    elif ndim == 1:
+        return T.vector(name=name, dtype=dtype)
+    elif ndim == 2:
+        return T.matrix(name=name, dtype=dtype)
+    elif ndim == 3:
+        return T.tensor3(name=name, dtype=dtype)
+    elif ndim == 4:
+        return T.tensor4(name=name, dtype=dtype)
+    else:
+        raise Exception('ndim too large: ' + str(ndim))
+
+
+def shape(x):
+    '''Return the shape of a tensor.
+
+    Warning: type returned will be different for
+    Theano backend (Theano tensor type) and TF backend (TF TensorShape).
+    '''
+    return x.shape
+
+
+def ndim(x):
+    return x.ndim
+
+
+def eval(x):
+    '''Run a graph.
+    '''
+    return x.eval()
+
+
+def zeros(shape, dtype=_FLOATX, name=None):
+    '''Instantiate an all-zeros variable.
+    '''
+    return variable(np.zeros(shape), dtype, name)
+
+
+def ones(shape, dtype=_FLOATX, name=None):
+    '''Instantiate an all-ones variable.
+    '''
+    return variable(np.ones(shape), dtype, name)
+
+
+def ones_like(x):
+    return T.ones_like(x)
+
+
+def zeros_like(x):
+    return T.zeros_like(x)
+
+
+def count_params(x):
+    '''Return number of scalars in a tensor.
+
+    Return: numpy integer.
+    '''
+    return np.prod(x.shape.eval())
+
+
+def cast(x, dtype):
+    return T.cast(x, dtype)
+
+
+# LINEAR ALGEBRA
+
+'''
+Assumed overridden:
++, -, /, *, +=, -=, *=, /=
+'''
+
+
+def dot(x, y):
+    return T.dot(x, y)
+
+
+def transpose(x):
+    return T.transpose(x)
+
+
+def gather(reference, indices):
+    '''reference: a tensor.
+    indices: an int tensor of indices.
+
+    Return: a tensor of same type as reference.
+    '''
+    return reference[indices]
+
+
+# ELEMENT-WISE OPERATIONS
+
+
+def max(x, axis=None, keepdims=False):
+    return T.max(x, axis=axis, keepdims=keepdims)
+
+
+def min(x, axis=None, keepdims=False):
+    return T.min(x, axis=axis, keepdims=keepdims)
+
+
+def sum(x, axis=None, keepdims=False):
+    '''Sum of the values in a tensor, alongside the specified axis.
+    '''
+    return T.sum(x, axis=axis, keepdims=keepdims)
+
+
+def prod(x, axis=None, keepdims=False):
+    '''Multiply the values in a tensor, alongside the specified axis.
+    '''
+    return T.prod(x, axis=axis, keepdims=keepdims)
+
+
+def mean(x, axis=None, keepdims=False):
+    return T.mean(x, axis=axis, keepdims=keepdims)
+
+
+def std(x, axis=None, keepdims=False):
+    return T.std(x, axis=axis, keepdims=keepdims)
+
+
+def any(x, axis=None, keepdims=False):
+    '''Bitwise reduction (logical OR).
+    '''
+    return T.any(x, axis=axis, keepdims=keepdims)
+
+
+def argmax(x, axis=-1):
+    return T.argmax(x, axis=axis, keepdims=False)
+
+
+def argmin(x, axis=-1):
+    return T.argmin(x, axis=axis, keepdims=False)
+
+
+def square(x):
+    return T.sqr(x)
+
+
+def abs(x):
+    return T.abs_(x)
+
+
+def sqrt(x):
+    x = T.clip(x, 0., np.inf)
+    return T.sqrt(x)
+
+
+def exp(x):
+    return T.exp(x)
+
+
+def log(x):
+    return T.log(x)
+
+
+def round(x):
+    return T.round(x)
+
+
+def pow(x, a):
+    return T.pow(x, a)
+
+
+def clip(x, min_value, max_value):
+    if max_value < min_value:
+        max_value = min_value
+    return T.clip(x, min_value, max_value)
+
+
+def equal(x, y):
+    return T.eq(x, y)
+
+
+def maximum(x, y):
+    return T.maximum(x, y)
+
+
+def minimum(x, y):
+    return T.minimum(x, y)
+
+
+# SHAPE OPERATIONS
+
+def concatenate(tensors, axis=-1):
+    return T.concatenate(tensors, axis=axis)
+
+
+def reshape(x, shape):
+    return T.reshape(x, shape)
+
+
+def permute_dimensions(x, pattern):
+    '''Transpose dimensions.
+
+    pattern should be a tuple or list of
+    dimension indices, e.g. [0, 2, 1].
+    '''
+    pattern = tuple(pattern)
+    return x.dimshuffle(pattern)
+
+
+def repeat(x, n):
+    '''Repeat a 2D tensor:
+
+    if x has shape (samples, dim) and n=2,
+    the output will have shape (samples, 2, dim)
+    '''
+    tensors = [x] * n
+    stacked = T.stack(*tensors)
+    return stacked.dimshuffle((1, 0, 2))
+
+
+def tile(x, n):
+    return T.tile(x, n)
+
+
+def flatten(x):
+    '''Turn a n-D tensor into a 2D tensor where
+    the first dimension is conserved.
+    '''
+    x = T.reshape(x, (x.shape[0], T.prod(x.shape) // x.shape[0]))
+    return x
+
+
+def expand_dims(x, dim=-1):
+    '''Add a 1-sized dimension at index "dim".
+    '''
+    pattern = [i for i in range(x.type.ndim)]
+    if dim < 0:
+        if x.type.ndim == 0:
+            dim = 0
+        else:
+            dim = dim % x.type.ndim + 1
+    pattern.insert(dim, 'x')
+    return x.dimshuffle(pattern)
+
+
+def squeeze(x, axis):
+    '''Remove a 1-dimension from the tensor at index "axis".
+    '''
+    x = T.addbroadcast(x, axis)
+    return T.squeeze(x)
+
+
+def temporal_padding(x, padding=1):
+    '''Pad the middle dimension of a 3D tensor
+    with "padding" zeros left and right.
+
+    Appologies for the inane API, but Theano makes this
+    really hard.
+    '''
+    input_shape = x.shape
+    output_shape = (input_shape[0],
+                    input_shape[1] + 2 * padding,
+                    input_shape[2])
+    output = T.zeros(output_shape)
+    return T.set_subtensor(output[:, padding:x.shape[1] + padding, :], x)
+
+
+def spatial_2d_padding(x, padding=(1, 1), dim_ordering='th'):
+    '''Pad the 2nd and 3rd dimensions of a 4D tensor
+    with "padding[0]" and "padding[1]" (resp.) zeros left and right.
+    '''
+    input_shape = x.shape
+    if dim_ordering == 'th':
+        output_shape = (input_shape[0],
+                        input_shape[1],
+                        input_shape[2] + 2 * padding[0],
+                        input_shape[3] + 2 * padding[1])
+        output = T.zeros(output_shape)
+        indices = (slice(None),
+                   slice(None),
+                   slice(padding[0], input_shape[2] + padding[0]),
+                   slice(padding[1], input_shape[3] + padding[1]))
+
+    elif dim_ordering == 'tf':
+        output_shape = (input_shape[0],
+                        input_shape[1] + 2 * padding[0],
+                        input_shape[2] + 2 * padding[1],
+                        input_shape[3])
+        output = T.zeros(output_shape)
+        indices = (slice(None),
+                   slice(padding[0], input_shape[1] + padding[0]),
+                   slice(padding[1], input_shape[2] + padding[1]),
+                   slice(None))
+    else:
+        raise Exception('Invalid dim_ordering: ' + dim_ordering)
+    return T.set_subtensor(output[indices], x)
+
+# VALUE MANIPULATION
+
+
+def get_value(x):
+    if not hasattr(x, 'get_value'):
+        raise Exception("'get_value() can only be called on a variable. " +
+                        "If you have an expression instead, use eval().")
+    return x.get_value()
+
+
+def set_value(x, value):
+    x.set_value(np.asarray(value, dtype=x.dtype))
+
+
+# GRAPH MANIPULATION
+
+class Function(object):
+
+    def __init__(self, inputs, outputs, updates=[], **kwargs):
+        self.function = theano.function(inputs, outputs, updates=updates,
+                                        allow_input_downcast=True, **kwargs)
+
+    def __call__(self, inputs):
+        return self.function(*inputs)
+
+
+def function(inputs, outputs, updates=[]):
+    return Function(inputs, outputs, updates=updates)
+
+
+def gradients(loss, variables):
+    return T.grad(loss, variables)
+
+
+# CONTROL FLOW
+
+def rnn(step_function, inputs, initial_states,
+        go_backwards=False, masking=True):
+    '''Iterates over the time dimension of a tensor.
+
+    Parameters
+    ----------
+    inputs: tensor of temporal data of shape (samples, time, ...)
+        (at least 3D).
+    step_function:
+        Parameters:
+            input: tensor with shape (samples, ...) (no time dimension),
+                representing input for the batch of samples at a certain
+                time step.
+            states: list of tensors.
+        Returns:
+            output: tensor with shape (samples, ...) (no time dimension),
+            new_states: list of tensors, same length and shapes
+                as 'states'.
+    initial_states: tensor with shape (samples, ...) (no time dimension),
+        containing the initial values for the states used in
+        the step function.
+    go_backwards: boolean. If True, do the iteration over
+        the time dimension in reverse order.
+    masking: boolean. If true, any input timestep inputs[s, i]
+        that is all-zeros will be skipped (states will be passed to
+        the next step unchanged) and the corresponding output will
+        be all zeros.
+
+    Returns
+    -------
+    A tuple (last_output, outputs, new_states).
+        last_output: the latest output of the rnn, of shape (samples, ...)
+        outputs: tensor with shape (samples, time, ...) where each
+            entry outputs[s, t] is the output of the step function
+            at time t for sample s.
+        new_states: list of tensors, latest states returned by
+            the step function, of shape (samples, ...).
+    '''
+    inputs = inputs.dimshuffle((1, 0, 2))
+
+    def _step(*args):
+        global single_result
+        input = args[0]
+        states = args[1:]
+        output, new_states = step_function(input, states)
+        if masking:
+            # if all-zero input timestep, return
+            # all-zero output and unchanged states
+            switch = T.any(input)
+            output = T.switch(switch, output, 0. * output)
+            return_states = []
+            for state, new_state in zip(states, new_states):
+                return_states.append(T.switch(switch, new_state, state))
+            return [output] + return_states
+        else:
+            return [output] + new_states
+
+    results, _ = theano.scan(
+        _step,
+        sequences=inputs,
+        outputs_info=[None] + initial_states,
+        go_backwards=go_backwards)
+
+    # deal with Theano API inconsistency
+    if type(results) is list:
+        outputs = results[0]
+        states = results[1:]
+    else:
+        outputs = results
+        states = []
+
+    outputs = T.squeeze(outputs)
+    last_output = outputs[-1]
+
+    outputs = outputs.dimshuffle((1, 0, 2))
+    states = [T.squeeze(state[-1]) for state in states]
+    return last_output, outputs, states
+
+
+def switch(condition, then_expression, else_expression):
+    '''condition: scalar tensor.
+    '''
+    return T.switch(condition, then_expression, else_expression)
+
+
+# NN OPERATIONS
+
+def relu(x, alpha=0., max_value=None):
+    x = T.nnet.relu(x, alpha)
+    if max_value is not None:
+        x = T.minimum(x, max_value)
+    return x
+
+
+def softmax(x):
+    return T.nnet.softmax(x)
+
+
+def softplus(x):
+    return T.nnet.softplus(x)
+
+
+def categorical_crossentropy(output, target, from_logits=False):
+    if from_logits:
+        output = T.nnet.softmax(output)
+    else:
+        # scale preds so that the class probas of each sample sum to 1
+        output /= output.sum(axis=-1, keepdims=True)
+    # avoid numerical instability with _EPSILON clipping
+    output = T.clip(output, _EPSILON, 1.0 - _EPSILON)
+    return T.nnet.categorical_crossentropy(output, target)
+
+
+def binary_crossentropy(output, target, from_logits=False):
+    if from_logits:
+        output = T.nnet.sigmoid(output)
+    # avoid numerical instability with _EPSILON clipping
+    output = T.clip(output, _EPSILON, 1.0 - _EPSILON)
+    return T.nnet.binary_crossentropy(output, target)
+
+
+def sigmoid(x):
+    return T.nnet.sigmoid(x)
+
+
+def hard_sigmoid(x):
+    return T.nnet.hard_sigmoid(x)
+
+
+def tanh(x):
+    return T.tanh(x)
+
+
+def dropout(x, level, seed=None):
+    if level < 0. or level >= 1:
+        raise Exception('Dropout level must be in interval [0, 1[.')
+    if seed is None:
+        seed = np.random.randint(10e6)
+    rng = RandomStreams(seed=seed)
+    retain_prob = 1. - level
+    x *= rng.binomial(x.shape, p=retain_prob, dtype=x.dtype)
+    x /= retain_prob
+    return x
+
+
+# CONVOLUTIONS
+
+
+def conv2d(x, kernel, strides=(1, 1), border_mode='valid', dim_ordering='th'):
+    '''
+    Run on cuDNN if available.
+    border_mode: string, "same" or "valid".
+    '''
+    if dim_ordering not in {'th', 'tf'}:
+        raise Exception('Unknown dim_ordering ' + str(dim_ordering))
+
+    if dim_ordering == 'tf':
+        # TF uses the last dimension as channel dimension,
+        # instead of the 2nd one.
+        # TH input shape: (samples, input_depth, rows, cols)
+        # TF input shape: (samples, rows, cols, input_depth)
+        # TH kernel shape: (depth, input_depth, rows, cols)
+        # TF kernel shape: (rows, cols, input_depth, depth)
+        x = x.dimshuffle((0, 3, 1, 2))
+        kernel = kernel.dimshuffle((3, 2, 0, 1))
+
+    if _on_gpu() and dnn.dnn_available():
+        if border_mode == 'same':
+            assert(strides == (1, 1))
+            pad_x = (kernel.shape[2] - strides[0]) // 2
+            pad_y = (kernel.shape[3] - strides[1]) // 2
+            conv_out = dnn.dnn_conv(img=x,
+                                    kerns=kernel,
+                                    border_mode=(pad_x, pad_y))
+        else:
+            conv_out = dnn.dnn_conv(img=x,
+                                    kerns=kernel,
+                                    border_mode=border_mode,
+                                    subsample=strides)
+    else:
+        if border_mode == 'same':
+            th_border_mode = 'full'
+            assert(strides == (1, 1))
+        elif border_mode == 'valid':
+            th_border_mode = 'valid'
+        else:
+            raise Exception('Border mode not supported: ' + str(border_mode))
+
+        conv_out = T.nnet.conv.conv2d(x, kernel,
+                                      border_mode=th_border_mode,
+                                      subsample=strides)
+        if border_mode == 'same':
+            shift_x = (kernel.shape[2] - 1) // 2
+            shift_y = (kernel.shape[3] - 1) // 2
+            conv_out = conv_out[:, :,
+                                shift_x:x.shape[2] + shift_x,
+                                shift_y:x.shape[3] + shift_y]
+    if dim_ordering == 'tf':
+        conv_out = conv_out.dimshuffle((0, 2, 3, 1))
+    return conv_out
+
+
+def maxpool2d(x, pool_size, strides=(1, 1), border_mode='valid',
+              dim_ordering='th'):
+    if border_mode == 'same':
+        # TODO: add implementation for border_mode="same"
+        raise Exception('border_mode="same" not supported with Theano.')
+    elif border_mode == 'valid':
+        ignore_border = True
+        padding = (0, 0)
+    else:
+        raise Exception('Invalid border mode: ' + str(border_mode))
+
+    if dim_ordering not in {'th', 'tf'}:
+        raise Exception('Unknown dim_ordering ' + str(dim_ordering))
+
+    if dim_ordering == 'tf':
+        x = x.dimshuffle((0, 3, 1, 2))
+
+    pool_out = downsample.max_pool_2d(x,
+                                      ds=pool_size,
+                                      st=strides,
+                                      ignore_border=ignore_border,
+                                      padding=padding,
+                                      mode='average_exc_pad')
+    if dim_ordering == 'tf':
+        pool_out = pool_out.dimshuffle((0, 2, 3, 1))
+    return pool_out
+
+
+# RANDOMNESS
+
+def random_normal(shape, mean=0.0, std=1.0, dtype=_FLOATX, seed=None):
+    if seed is None:
+        seed = np.random.randint(10e6)
+    rng = RandomStreams(seed=seed)
+    return rng.normal(size=shape, avg=mean, std=std, dtype=dtype)
+
+
+def random_uniform(shape, low=0.0, high=1.0, dtype=_FLOATX, seed=None):
+    if seed is None:
+        seed = np.random.randint(10e6)
+    rng = RandomStreams(seed=seed)
+    return rng.uniform(shape, low=low, high=high, dtype=dtype)
+
+
+
+'''
+more TODO:
+
+tensordot -> soon to be introduced in TF
+batched_tensordot -> reimplement
+'''

keras/callbacks.py

@@ -2,7 +2,9 @@ from __future__ import absolute_import
 from __future__ import print_function
 
 import numpy as np
-import time, json, warnings
+import time
+import json
+import warnings
 
 from collections import deque
 from .utils.generic_utils import Progbar
@@ -99,10 +101,11 @@ class Callback(object):
 class BaseLogger(Callback):
     def on_train_begin(self, logs={}):
         self.verbose = self.params['verbose']
+        self.nb_epoch = self.params['nb_epoch']
 
     def on_epoch_begin(self, epoch, logs={}):
         if self.verbose:
-            print('Epoch %d' % epoch)
+            print('Epoch %d/%d' % (epoch + 1, self.nb_epoch))
             self.progbar = Progbar(target=self.params['nb_sample'],
                                    verbose=self.verbose)
         self.seen = 0
@@ -172,34 +175,52 @@ class History(Callback):
 
 
 class ModelCheckpoint(Callback):
-    def __init__(self, filepath, monitor='val_loss', verbose=0, save_best_only=False):
+    def __init__(self, filepath, monitor='val_loss', verbose=0, save_best_only=False, mode='auto'):
 
         super(Callback, self).__init__()
         self.monitor = monitor
         self.verbose = verbose
         self.filepath = filepath
         self.save_best_only = save_best_only
-        self.best = np.Inf
+        
+        if mode not in ['auto', 'min', 'max']:
+            warnings.warn("ModelCheckpoint mode %s is unknown, fallback to auto mode" % (self.mode), RuntimeWarning)
+            mode = 'auto'
+            
+        if mode == "min":
+            self.monitor_op = np.less
+            self.best = np.Inf
+        elif mode == "max":
+            self.monitor_op = np.greater
+            self.best = -np.Inf
+        else:
+            if "acc" in self.monitor:
+                self.monitor_op = np.greater
+                self.best = -np.Inf
+            else:
+                self.monitor_op = np.less
+                self.best = np.Inf
 
     def on_epoch_end(self, epoch, logs={}):
+        filepath = self.filepath.format(epoch=epoch, **logs)
         if self.save_best_only:
             current = logs.get(self.monitor)
             if current is None:
                 warnings.warn("Can save best model only with %s available, skipping." % (self.monitor), RuntimeWarning)
             else:
-                if current < self.best:
+                if self.monitor_op(current, self.best):
                     if self.verbose > 0:
                         print("Epoch %05d: %s improved from %0.5f to %0.5f, saving model to %s"
-                              % (epoch, self.monitor, self.best, current, self.filepath))
+                              % (epoch, self.monitor, self.best, current, filepath))
                     self.best = current
-                    self.model.save_weights(self.filepath, overwrite=True)
+                    self.model.save_weights(filepath, overwrite=True)
                 else:
                     if self.verbose > 0:
                         print("Epoch %05d: %s did not improve" % (epoch, self.monitor))
         else:
             if self.verbose > 0:
-                print("Epoch %05d: saving model to %s" % (epoch, self.filepath))
-            self.model.save_weights(self.filepath, overwrite=True)
+                print("Epoch %05d: saving model to %s" % (epoch, filepath))
+            self.model.save_weights(filepath, overwrite=True)
 
 
 class EarlyStopping(Callback):
@@ -271,4 +292,4 @@ class LearningRateScheduler(Callback):
         self.schedule = schedule
 
     def on_epoch_begin(self, epoch, logs={}):
-        model.lr.set_value(self.schedule(epoch))
+        self.model.optimizer.lr.set_value(self.schedule(epoch))

keras/constraints.py

@@ -1,7 +1,5 @@
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
-import numpy as np
+from . import backend as K
 
 
 class Constraint(object):
@@ -17,8 +15,8 @@ class MaxNorm(Constraint):
         self.m = m
 
     def __call__(self, p):
-        norms = T.sqrt(T.sum(T.sqr(p), axis=0))
-        desired = T.clip(norms, 0, self.m)
+        norms = K.sqrt(K.sum(K.square(p), axis=0))
+        desired = K.clip(norms, 0, self.m)
         p = p * (desired / (1e-7 + norms))
         return p
 
@@ -29,13 +27,13 @@ class MaxNorm(Constraint):
 
 class NonNeg(Constraint):
     def __call__(self, p):
-        p *= T.ge(p, 0)
+        p *= K.cast(p >= 0., K.floatx())
         return p
 
 
 class UnitNorm(Constraint):
     def __call__(self, p):
-        return p / T.sqrt(T.sum(p**2, axis=-1, keepdims=True))
+        return p / K.sqrt(K.sum(K.square(p), axis=-1, keepdims=True))
 
 identity = Constraint
 maxnorm = MaxNorm

keras/datasets/cifar.py

@@ -1,15 +1,15 @@
 # -*- coding: utf-8 -*-
 from __future__ import absolute_import
 import sys
-import six.moves.cPickle
+from six.moves import cPickle
 from six.moves import range
 
 def load_batch(fpath, label_key='labels'):
     f = open(fpath, 'rb')
     if sys.version_info < (3,):
-        d = six.moves.cPickle.load(f)
+        d = cPickle.load(f)
     else:
-        d = six.moves.cPickle.load(f, encoding="bytes")
+        d = cPickle.load(f, encoding="bytes")
         # decode utf8
         for k, v in d.items():
             del(d[k])

keras/datasets/data_utils.py

@@ -1,14 +1,17 @@
 from __future__ import absolute_import
 from __future__ import print_function
 
-import tarfile, inspect, os
+import tarfile
+import os
 from six.moves.urllib.request import FancyURLopener
 
 from ..utils.generic_utils import Progbar
 
+
 class ParanoidURLopener(FancyURLopener):
-  def http_error_default(self, url, fp, errcode, errmsg, headers):
-    raise Exception('URL fetch failure on {}: {} -- {}'.format(url, errcode, errmsg))
+    def http_error_default(self, url, fp, errcode, errmsg, headers):
+        raise Exception('URL fetch failure on {}: {} -- {}'.format(url, errcode, errmsg))
+
 
 def get_file(fname, origin, untar=False):
     datadir = os.path.expanduser(os.path.join('~', '.keras', 'datasets'))
@@ -21,11 +24,8 @@ def get_file(fname, origin, untar=False):
     else:
         fpath = os.path.join(datadir, fname)
 
-    try:
-        f = open(fpath)
-    except:
+    if not os.path.exists(fpath):
         print('Downloading data from',  origin)
-
         global progbar
         progbar = None
 

keras/datasets/imdb.py

@@ -1,5 +1,5 @@
 from __future__ import absolute_import
-import six.moves.cPickle
+from six.moves import cPickle
 import gzip
 from .data_utils import get_file
 import random
@@ -17,7 +17,7 @@ def load_data(path="imdb.pkl", nb_words=None, skip_top=0, maxlen=None, test_spli
     else:
         f = open(path, 'rb')
 
-    X, labels = six.moves.cPickle.load(f)
+    X, labels = cPickle.load(f)
     f.close()
 
     np.random.seed(seed)

keras/datasets/mnist.py

@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 import gzip
 from .data_utils import get_file
-import six.moves.cPickle
+from six.moves import cPickle
 import sys
 
 
@@ -14,9 +14,9 @@ def load_data(path="mnist.pkl.gz"):
         f = open(path, 'rb')
 
     if sys.version_info < (3,):
-        data = six.moves.cPickle.load(f)
+        data = cPickle.load(f)
     else:
-        data = six.moves.cPickle.load(f, encoding="bytes")
+        data = cPickle.load(f, encoding="bytes")
 
     f.close()
 

keras/datasets/reuters.py

@@ -5,7 +5,7 @@ from .data_utils import get_file
 import string
 import random
 import os
-import six.moves.cPickle
+from six.moves import cPickle
 from six.moves import zip
 import numpy as np
 
@@ -78,8 +78,8 @@ def make_reuters_dataset(path=os.path.join('datasets', 'temp', 'reuters21578'),
     dataset = (X, labels)
     print('-')
     print('Saving...')
-    six.moves.cPickle.dump(dataset, open(os.path.join('datasets', 'data', 'reuters.pkl'), 'w'))
-    six.moves.cPickle.dump(tokenizer.word_index, open(os.path.join('datasets', 'data', 'reuters_word_index.pkl'), 'w'))
+    cPickle.dump(dataset, open(os.path.join('datasets', 'data', 'reuters.pkl'), 'w'))
+    cPickle.dump(tokenizer.word_index, open(os.path.join('datasets', 'data', 'reuters_word_index.pkl'), 'w'))
 
 
 def load_data(path="reuters.pkl", nb_words=None, skip_top=0, maxlen=None, test_split=0.2, seed=113,
@@ -88,7 +88,7 @@ def load_data(path="reuters.pkl", nb_words=None, skip_top=0, maxlen=None, test_s
     path = get_file(path, origin="https://s3.amazonaws.com/text-datasets/reuters.pkl")
     f = open(path, 'rb')
 
-    X, labels = six.moves.cPickle.load(f)
+    X, labels = cPickle.load(f)
     f.close()
 
     np.random.seed(seed)
@@ -140,7 +140,7 @@ def load_data(path="reuters.pkl", nb_words=None, skip_top=0, maxlen=None, test_s
 def get_word_index(path="reuters_word_index.pkl"):
     path = get_file(path, origin="https://s3.amazonaws.com/text-datasets/reuters_word_index.pkl")
     f = open(path, 'rb')
-    return six.moves.cPickle.load(f)
+    return cPickle.load(f)
 
 
 if __name__ == "__main__":

keras/initializations.py

@@ -1,9 +1,6 @@
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
 import numpy as np
-
-from .utils.theano_utils import sharedX, shared_zeros, shared_ones
+from . import backend as K
 
 
 def get_fans(shape):
@@ -13,11 +10,11 @@ def get_fans(shape):
 
 
 def uniform(shape, scale=0.05):
-    return sharedX(np.random.uniform(low=-scale, high=scale, size=shape))
+    return K.variable(np.random.uniform(low=-scale, high=scale, size=shape))
 
 
 def normal(shape, scale=0.05):
-    return sharedX(np.random.randn(*shape) * scale)
+    return K.variable(np.random.randn(*shape) * scale)
 
 
 def lecun_uniform(shape):
@@ -58,7 +55,7 @@ def he_uniform(shape):
 
 
 def orthogonal(shape, scale=1.1):
-    ''' From Lasagne
+    ''' From Lasagne. Reference: Saxe et al., http://arxiv.org/abs/1312.6120
     '''
     flat_shape = (shape[0], np.prod(shape[1:]))
     a = np.random.normal(0.0, 1.0, flat_shape)
@@ -66,22 +63,22 @@ def orthogonal(shape, scale=1.1):
     # pick the one with the correct shape
     q = u if u.shape == flat_shape else v
     q = q.reshape(shape)
-    return sharedX(scale * q[:shape[0], :shape[1]])
+    return K.variable(scale * q[:shape[0], :shape[1]])
 
 
 def identity(shape, scale=1):
     if len(shape) != 2 or shape[0] != shape[1]:
         raise Exception("Identity matrix initialization can only be used for 2D square matrices")
     else:
-        return sharedX(scale * np.identity(shape[0]))
+        return K.variable(scale * np.identity(shape[0]))
 
 
 def zero(shape):
-    return shared_zeros(shape)
+    return K.zeros(shape)
 
 
 def one(shape):
-    return shared_ones(shape)
+    return K.ones(shape)
 
 
 from .utils.generic_utils import get_from_module

keras/layers/advanced_activations.py

@@ -1,50 +1,75 @@
 from .. import initializations
 from ..layers.core import Layer, MaskedLayer
-from ..utils.theano_utils import shared_zeros, shared_ones, sharedX
-import theano.tensor as T
+from .. import backend as K
 import numpy as np
 
 
 class LeakyReLU(MaskedLayer):
-    def __init__(self, alpha=0.3):
-        super(LeakyReLU, self).__init__()
+    def __init__(self, alpha=0.3, **kwargs):
+        super(LeakyReLU, self).__init__(**kwargs)
         self.alpha = alpha
 
     def get_output(self, train):
         X = self.get_input(train)
-        return ((X + abs(X)) / 2.0) + self.alpha * ((X - abs(X)) / 2.0)
+        return K.relu(X, alpha=self.alpha)
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "alpha": self.alpha}
+        config = {"name": self.__class__.__name__,
+                  "alpha": self.alpha}
+        base_config = super(LeakyReLU, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class PReLU(MaskedLayer):
     '''
         Reference:
-            Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
+            Delving Deep into Rectifiers: Surpassing Human-Level
+            Performance on ImageNet Classification
                 http://arxiv.org/pdf/1502.01852v1.pdf
     '''
-    def __init__(self, input_shape, init='zero', weights=None):
-        super(PReLU, self).__init__()
+    def __init__(self, init='zero', weights=None, **kwargs):
         self.init = initializations.get(init)
+        self.initial_weights = weights
+        super(PReLU, self).__init__(**kwargs)
+
+    def build(self):
+        input_shape = self.input_shape[1:]
         self.alphas = self.init(input_shape)
         self.params = [self.alphas]
-        self.input_shape = input_shape
 
-        if weights is not None:
-            self.set_weights(weights)
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
 
     def get_output(self, train):
         X = self.get_input(train)
-        pos = ((X + abs(X)) / 2.0)
-        neg = self.alphas * ((X - abs(X)) / 2.0)
+        pos = K.relu(X)
+        neg = self.alphas * (X - abs(X)) * 0.5
         return pos + neg
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_shape": self.input_shape,
-                "init": self.init.__name__}
+        config = {"name": self.__class__.__name__,
+                  "init": self.init.__name__}
+        base_config = super(PReLU, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
+
+
+class ELU(MaskedLayer):
+    def __init__(self, alpha=1.0, **kwargs):
+        super(ELU, self).__init__(**kwargs)
+        self.alpha = alpha
+
+    def get_output(self, train):
+        X = self.get_input(train)
+        pos = K.relu(X)
+        neg = (X - abs(X)) * 0.5
+        return pos + self.alpha * (K.exp(neg) - 1.)
+
+    def get_config(self):
+        config = {"name": self.__class__.__name__,
+                  "alpha": self.alpha}
+        base_config = super(ELU, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class ParametricSoftplus(MaskedLayer):
@@ -52,31 +77,38 @@ class ParametricSoftplus(MaskedLayer):
         Parametric Softplus of the form: alpha * log(1 + exp(beta * X))
 
         Reference:
-            Inferring Nonlinear Neuronal Computation Based on Physiologically Plausible Inputs
+            Inferring Nonlinear Neuronal Computation
+            Based on Physiologically Plausible Inputs
             http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1003143
     '''
-    def __init__(self, input_shape, alpha_init=0.2, beta_init=5.0, weights=None):
-
-        super(ParametricSoftplus, self).__init__()
+    def __init__(self, alpha_init=0.2, beta_init=5.0,
+                 weights=None, **kwargs):
         self.alpha_init = alpha_init
         self.beta_init = beta_init
-        self.alphas = sharedX(alpha_init * np.ones(input_shape))
-        self.betas = sharedX(beta_init * np.ones(input_shape))
-        self.params = [self.alphas, self.betas]
-        self.input_shape = input_shape
+        self.initial_weights = weights
+        super(ParametricSoftplus, self).__init__(**kwargs)
 
-        if weights is not None:
-            self.set_weights(weights)
+    def build(self):
+        input_shape = self.input_shape[1:]
+        self.alphas = K.variable(self.alpha_init * np.ones(input_shape))
+        self.betas = K.variable(self.beta_init * np.ones(input_shape))
+        self.params = [self.alphas, self.betas]
+
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
 
     def get_output(self, train):
         X = self.get_input(train)
-        return T.nnet.softplus(self.betas * X) * self.alphas
+        return K.softplus(self.betas * X) * self.alphas
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_shape": self.input_shape,
-                "alpha_init": self.alpha_init,
-                "beta_init": self.beta_init}
+        config = {"name": self.__class__.__name__,
+                  "alpha_init": self.alpha_init,
+                  "beta_init": self.beta_init}
+        base_config = super(ParametricSoftplus, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
+
 
 class ThresholdedLinear(MaskedLayer):
     '''
@@ -86,19 +118,22 @@ class ThresholdedLinear(MaskedLayer):
             Zero-Bias Autoencoders and the Benefits of Co-Adapting Features
             http://arxiv.org/pdf/1402.3337.pdf
     '''
-    def __init__(self, theta=1.0):
-        super(ThresholdedLinear, self).__init__()
+    def __init__(self, theta=1.0, **kwargs):
+        super(ThresholdedLinear, self).__init__(**kwargs)
         self.theta = theta
-    
+
     def get_output(self, train):
         X = self.get_input(train)
-        return T.switch( abs(X) < self.theta, 0, X )
+        return K.switch(K.abs(X) < self.theta, 0, X)
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-            "theta": self.theta}
+        config = {"name": self.__class__.__name__,
+                  "theta": self.theta}
+        base_config = super(ThresholdedLinear, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
-class ThresholdedReLu(MaskedLayer):
+
+class ThresholdedReLU(MaskedLayer):
     '''
         Thresholded Rectified Activation
 
@@ -106,14 +141,16 @@ class ThresholdedReLu(MaskedLayer):
             Zero-Bias Autoencoders and the Benefits of Co-Adapting Features
             http://arxiv.org/pdf/1402.3337.pdf
     '''
-    def __init__(self, theta=1.0):
-        super(ThresholdedReLu, self).__init__()
+    def __init__(self, theta=1.0, **kwargs):
+        super(ThresholdedReLU, self).__init__(**kwargs)
         self.theta = theta
-    
+
     def get_output(self, train):
         X = self.get_input(train)
-        return T.switch( X > self.theta, X, 0 )
+        return K.switch(X > self.theta, X, 0)
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-            "theta": self.theta}
+        config = {"name": self.__class__.__name__,
+                  "theta": self.theta}
+        base_config = super(ThresholdedReLU, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/layers/containers.py

@@ -2,9 +2,9 @@
 from __future__ import absolute_import
 from __future__ import print_function
 
-import theano.tensor as T
-from ..layers.core import Layer, Merge
-from ..utils.theano_utils import ndim_tensor
+from collections import OrderedDict
+from .. import backend as K
+from ..layers.core import Layer, Merge, Siamese, SiameseHead
 from six.moves import range
 
 
@@ -20,11 +20,6 @@ class Sequential(Layer):
 
     def __init__(self, layers=[]):
         self.layers = []
-        self.params = []
-        self.regularizers = []
-        self.constraints = []
-        self.updates = []
-
         for layer in layers:
             self.add(layer)
 
@@ -37,13 +32,42 @@ class Sequential(Layer):
             self.layers[-1].set_previous(self.layers[-2])
             if not hasattr(self.layers[0], 'input'):
                 self.set_input()
-        layer.init_updates()
 
-        params, regularizers, constraints, updates = layer.get_params()
-        self.params += params
-        self.regularizers += regularizers
-        self.constraints += constraints
-        self.updates += updates
+    @property
+    def params(self):
+        params = []
+        for l in self.layers:
+            if l.trainable:
+                params += l.get_params()[0]
+        return params
+
+    @property
+    def regularizers(self):
+        regularizers = []
+        for l in self.layers:
+            if l.trainable:
+                regularizers += l.get_params()[1]
+        return regularizers
+
+    @property
+    def constraints(self):
+        constraints = []
+        for l in self.layers:
+            if l.trainable:
+                constraints += l.get_params()[2]
+        return constraints
+
+    @property
+    def updates(self):
+        updates = []
+        for l in self.layers:
+            if l.trainable:
+                updates += l.get_params()[3]
+        return updates
+
+    @property
+    def output_shape(self):
+        return self.layers[-1].output_shape
 
     def get_output(self, train=False):
         return self.layers[-1].get_output(train)
@@ -51,8 +75,8 @@ class Sequential(Layer):
     def set_input(self):
         for l in self.layers:
             if hasattr(l, 'input'):
-                ndim = l.input.ndim
-                self.layers[0].input = ndim_tensor(ndim)
+                ndim = len(K.get_shape(l.input))
+                self.layers[0].input = K.placeholder(ndim=ndim)
                 break
 
     def get_input(self, train=False):
@@ -60,6 +84,10 @@ class Sequential(Layer):
             self.set_input()
         return self.layers[0].get_input(train)
 
+    @property
+    def input_shape(self):
+        return self.layers[0].input_shape
+
     @property
     def input(self):
         return self.get_input()
@@ -80,6 +108,9 @@ class Sequential(Layer):
         return {"name": self.__class__.__name__,
                 "layers": [layer.get_config() for layer in self.layers]}
 
+    def count_params(self):
+        return sum([layer.count_params() for layer in self.layers])
+
 
 class Graph(Layer):
     '''
@@ -91,7 +122,6 @@ class Graph(Layer):
         when it has exactly one input and one output.
 
         inherited from Layer:
-            - get_params
             - get_output_mask
             - supports_masked_input
             - get_weights
@@ -99,7 +129,7 @@ class Graph(Layer):
     '''
     def __init__(self):
         self.namespace = set()  # strings
-        self.nodes = {}  # layer-like
+        self.nodes = OrderedDict()  # layer-like
         self.inputs = {}  # layer-like
         self.input_order = []  # strings
         self.outputs = {}  # layer-like
@@ -108,11 +138,6 @@ class Graph(Layer):
         self.output_config = []  # dicts
         self.node_config = []  # dicts
 
-        self.params = []
-        self.regularizers = []
-        self.constraints = []
-        self.updates = []
-
     @property
     def nb_input(self):
         return len(self.inputs)
@@ -121,6 +146,38 @@ class Graph(Layer):
     def nb_output(self):
         return len(self.outputs)
 
+    @property
+    def params(self):
+        params = []
+        for l in self.nodes.values():
+            if l.trainable:
+                params += l.get_params()[0]
+        return params
+
+    @property
+    def regularizers(self):
+        regularizers = []
+        for l in self.nodes.values():
+            if l.trainable:
+                regularizers += l.get_params()[1]
+        return regularizers
+
+    @property
+    def constraints(self):
+        constraints = []
+        for l in self.nodes.values():
+            if l.trainable:
+                constraints += l.get_params()[2]
+        return constraints
+
+    @property
+    def updates(self):
+        updates = []
+        for l in self.nodes.values():
+            if l.trainable:
+                updates += l.get_params()[3]
+        return updates
+
     def set_previous(self, layer, connection_map={}):
         if self.nb_input != layer.nb_output:
             raise Exception('Cannot connect layers: input count does not match output count.')
@@ -145,32 +202,45 @@ class Graph(Layer):
     def input(self):
         return self.get_input()
 
+    @property
+    def output_shape(self):
+        if self.nb_output == 1:
+            # return tuple
+            return self.outputs[self.output_order[0]].output_shape
+        else:
+            # return dictionary mapping output names to shape tuples
+            return dict([(k, v.output_shape) for k, v in self.outputs.items()])
+
     def get_output(self, train=False):
         if len(self.inputs) == len(self.outputs) == 1:
             return self.outputs[self.output_order[0]].get_output(train)
         else:
             return dict([(k, v.get_output(train)) for k, v in self.outputs.items()])
 
-    def add_input(self, name, ndim=2, dtype='float'):
+    def add_input(self, name, input_shape, dtype='float'):
         if name in self.namespace:
             raise Exception('Duplicate node identifier: ' + name)
         self.namespace.add(name)
         self.input_order.append(name)
         layer = Layer()  # empty layer
+        layer.set_input_shape(input_shape)
         if dtype == 'float':
-            layer.input = ndim_tensor(ndim)
+            layer.input = K.placeholder(shape=layer.input_shape, name=name)
         else:
-            if ndim == 2:
-                layer.input = T.imatrix()
+            if len(input_shape) == 1:
+                layer.input = K.placeholder(shape=layer.input_shape,
+                                            dtype='int32',
+                                            name=name)
             else:
                 raise Exception('Type "int" can only be used with ndim==2 (Embedding).')
-        layer.input.name = name
         self.inputs[name] = layer
-        self.input_config.append({'name': name, 'ndim': ndim, 'dtype': dtype})
+        self.input_config.append({'name': name,
+                                  'input_shape': input_shape,
+                                  'dtype': dtype})
 
-    def add_node(self, layer, name, input=None, inputs=[], merge_mode='concat', create_output=False):
-        if hasattr(layer, 'set_name'):
-            layer.set_name(name)
+    def add_node(self, layer, name, input=None, inputs=[],
+                 merge_mode='concat', concat_axis=-1, dot_axes=-1,
+                 create_output=False):
         if name in self.namespace:
             raise Exception('Duplicate node identifier: ' + name)
         if input:
@@ -189,7 +259,8 @@ class Graph(Layer):
                     to_merge.append(self.inputs[n])
                 else:
                     raise Exception('Unknown identifier: ' + n)
-            merge = Merge(to_merge, mode=merge_mode)
+            merge = Merge(to_merge, mode=merge_mode,
+                          concat_axis=concat_axis, dot_axes=dot_axes)
             layer.set_previous(merge)
 
         self.namespace.add(name)
@@ -197,18 +268,91 @@ class Graph(Layer):
         self.node_config.append({'name': name,
                                  'input': input,
                                  'inputs': inputs,
-                                 'merge_mode': merge_mode})
-        layer.init_updates()
-        params, regularizers, constraints, updates = layer.get_params()
-        self.params += params
-        self.regularizers += regularizers
-        self.constraints += constraints
-        self.updates += updates
+                                 'merge_mode': merge_mode,
+                                 'concat_axis': concat_axis,
+                                 'dot_axes': dot_axes,
+                                 'create_output': create_output})
 
         if create_output:
             self.add_output(name, input=name)
 
-    def add_output(self, name, input=None, inputs=[], merge_mode='concat'):
+    def add_shared_node(self, layer, name, inputs=[], merge_mode=None,
+                        concat_axis=-1, dot_axes=-1, outputs=[],
+                        create_output=False):
+        '''
+        Used to shared / multi input-multi output node
+
+        Arguments
+        ------------
+        layer - The layer to be shared across multiple inputs
+        name - Name of the shared layer
+        inputs - List of names of input nodes
+        merge_mode - Similar to merge_mode argument of add_node()
+        concat_axis - Similar to concat_axis argument of add_node()
+        dot_axes - Similar to dot_axes argument of add_node()
+        outputs - Names for output nodes. Used when merge_mode = None
+        create_output -  Similar to create_output argument of add_node().
+            Output will be created only if merge_mode is given
+        '''
+        if name in self.namespace:
+            raise Exception('Duplicate node identifier: ' + name)
+        for o in outputs:
+            if o in self.namespace:
+                raise Exception('Duplicate node identifier: ' + o)
+        if merge_mode:
+            if merge_mode not in {'sum', 'ave', 'mul', 'dot', 'cos', 'concat', 'join'}:
+                raise Eception("Invalid merge mode")
+        layers = []
+        for i in range(len(inputs)):
+            input = inputs[i]
+            if input in self.nodes:
+                n = self.nodes[input]
+                if n.__class__.__name__ == 'Siamese':
+                    if n.merge_mode is None:
+                        for j in range(len(n.inputs)):
+                            sh = SiameseHead(j)
+                            sh.previous = n
+                            layers.append(sh)
+                    else:
+                        layers.append(n)
+                else:
+                    layers.append(n)
+            elif input in self.inputs:
+                n = self.inputs[input]
+                layers.append(n)
+            else:
+                raise Exception('Unknown identifier: ' + input)
+        s = Siamese(layer, layers, merge_mode, concat_axis=concat_axis, dot_axes=dot_axes)
+        s.set_name(name)
+        self.namespace.add(name)
+        self.nodes[name] = s
+        self.node_config.append({'name': name,
+                                 'inputs': inputs,
+                                 'merge_mode': merge_mode,
+                                 'concat_axis': concat_axis,
+                                 'dot_axes': dot_axes,
+                                 'create_output': create_output if merge_mode else False})
+        if not merge_mode:
+            for i in range(len(outputs)):
+                sh = SiameseHead(i)
+                sh.previous = s
+                sh_name = outputs[i]
+                sh.set_name(sh_name)
+                self.namespace.add(sh_name)
+                self.nodes[sh_name] = sh
+                self.node_config.append({'name': sh_name,
+                                         'inputs': [s],
+                                         'create_output': create_output})
+                if create_output:
+                    self.add_output(sh_name, input=sh_name)
+
+        if create_output and merge_mode:
+            if merge_mode == 'join':
+                raise Exception("Output can not be of type OrderedDict")
+            self.add_output(name, input=name)
+
+    def add_output(self, name, input=None, inputs=[],
+                   merge_mode='concat', concat_axis=-1, dot_axes=-1):
         if name in self.output_order:
             raise Exception('Duplicate output identifier: ' + name)
         if input:
@@ -224,14 +368,17 @@ class Graph(Layer):
                 if n not in self.nodes:
                     raise Exception('Unknown identifier: ' + n)
                 to_merge.append(self.nodes[n])
-            merge = Merge(to_merge, mode=merge_mode)
+            merge = Merge(to_merge, mode=merge_mode,
+                          concat_axis=concat_axis, dot_axes=dot_axes)
             self.outputs[name] = merge
 
         self.output_order.append(name)
         self.output_config.append({'name': name,
                                    'input': input,
                                    'inputs': inputs,
-                                   'merge_mode': merge_mode})
+                                   'merge_mode': merge_mode,
+                                   'concat_axis': concat_axis,
+                                   'dot_axes': dot_axes})
 
     def get_config(self):
         return {"name": self.__class__.__name__,
@@ -241,3 +388,18 @@ class Graph(Layer):
                 "input_order": self.input_order,
                 "output_order": self.output_order,
                 "nodes": dict([(c["name"], self.nodes[c["name"]].get_config()) for c in self.node_config])}
+
+    def count_params(self):
+        return sum([layer.count_params() for layer in self.nodes.values()])
+
+    def get_weights(self):
+        weights = []
+        for layer in self.nodes.values():
+            weights += layer.get_weights()
+        return weights
+
+    def set_weights(self, weights):
+        for layer in self.nodes.values():
+            nb_param = len(layer.get_weights())
+            layer.set_weights(weights[:nb_param])
+            weights = weights[nb_param:]

keras/layers/convolutional.py

@@ -1,294 +1,505 @@
 # -*- coding: utf-8 -*-
 from __future__ import absolute_import
 
-import theano
-import theano.tensor as T
-from theano.sandbox.cuda import dnn
-
+from .. import backend as K
 from .. import activations, initializations, regularizers, constraints
-from ..utils.theano_utils import shared_zeros
 from ..layers.core import Layer
 
 
+def conv_output_length(input_length, filter_size, border_mode, stride):
+    if input_length is None:
+        return None
+    assert border_mode in {'same', 'valid'}
+    if border_mode == 'same':
+        output_length = input_length
+    elif border_mode == 'valid':
+        output_length = input_length - filter_size + 1
+    return (output_length + stride - 1) // stride
+
+
 class Convolution1D(Layer):
-    def __init__(self, input_dim, nb_filter, filter_length,
+    input_ndim = 3
+
+    def __init__(self, nb_filter, filter_length,
                  init='uniform', activation='linear', weights=None,
                  border_mode='valid', subsample_length=1,
                  W_regularizer=None, b_regularizer=None, activity_regularizer=None,
-                 W_constraint=None, b_constraint=None):
+                 W_constraint=None, b_constraint=None,
+                 input_dim=None, input_length=None, **kwargs):
 
-        if border_mode not in {'valid', 'full', 'same'}:
+        if border_mode not in {'valid', 'same'}:
             raise Exception('Invalid border mode for Convolution1D:', border_mode)
-
-        super(Convolution1D, self).__init__()
         self.nb_filter = nb_filter
-        self.input_dim = input_dim
         self.filter_length = filter_length
-        self.subsample_length = subsample_length
         self.init = initializations.get(init)
         self.activation = activations.get(activation)
-        self.subsample = (1, subsample_length)
+        assert border_mode in {'valid', 'same'}, 'border_mode must be in {valid, same}'
         self.border_mode = border_mode
+        self.subsample_length = subsample_length
 
-        self.input = T.tensor3()
-        self.W_shape = (nb_filter, input_dim, filter_length, 1)
-        self.W = self.init(self.W_shape)
-        self.b = shared_zeros((nb_filter,))
-
-        self.params = [self.W, self.b]
-
-        self.regularizers = []
+        self.subsample = (subsample_length, 1)
 
         self.W_regularizer = regularizers.get(W_regularizer)
-        if self.W_regularizer:
-            self.W_regularizer.set_param(self.W)
-            self.regularizers.append(self.W_regularizer)
-
         self.b_regularizer = regularizers.get(b_regularizer)
-        if self.b_regularizer:
-            self.b_regularizer.set_param(self.b)
-            self.regularizers.append(self.b_regularizer)
-
         self.activity_regularizer = regularizers.get(activity_regularizer)
-        if self.activity_regularizer:
-            self.activity_regularizer.set_layer(self)
-            self.regularizers.append(self.activity_regularizer)
 
         self.W_constraint = constraints.get(W_constraint)
         self.b_constraint = constraints.get(b_constraint)
         self.constraints = [self.W_constraint, self.b_constraint]
 
-        if weights is not None:
-            self.set_weights(weights)
+        self.initial_weights = weights
 
-    def get_output(self, train):
+        self.input_dim = input_dim
+        self.input_length = input_length
+        if self.input_dim:
+            kwargs['input_shape'] = (self.input_length, self.input_dim)
+        self.input = K.placeholder(ndim=3)
+        super(Convolution1D, self).__init__(**kwargs)
+
+    def build(self):
+        input_dim = self.input_shape[2]
+        self.W_shape = (self.nb_filter, input_dim, self.filter_length, 1)
+        self.W = self.init(self.W_shape)
+        self.b = K.zeros((self.nb_filter,))
+        self.params = [self.W, self.b]
+        self.regularizers = []
+
+        if self.W_regularizer:
+            self.W_regularizer.set_param(self.W)
+            self.regularizers.append(self.W_regularizer)
+
+        if self.b_regularizer:
+            self.b_regularizer.set_param(self.b)
+            self.regularizers.append(self.b_regularizer)
+
+        if self.activity_regularizer:
+            self.activity_regularizer.set_layer(self)
+            self.regularizers.append(self.activity_regularizer)
+
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
+
+    @property
+    def output_shape(self):
+        length = conv_output_length(self.input_shape[1],
+                                    self.filter_length,
+                                    self.border_mode,
+                                    self.subsample[0])
+        return (self.input_shape[0], length, self.nb_filter)
+
+    def get_output(self, train=False):
         X = self.get_input(train)
-        X = T.reshape(X, (X.shape[0], X.shape[1], X.shape[2], 1)).dimshuffle(0, 2, 1, 3)
+        X = K.expand_dims(X, -1)  # add a dimension of the right
+        X = K.permute_dimensions(X, (0, 2, 1, 3))
+        conv_out = K.conv2d(X, self.W, strides=self.subsample,
+                            border_mode=self.border_mode, dim_ordering='th')
 
-        border_mode = self.border_mode
-        if border_mode == 'same':
-            border_mode = 'full'
-
-        conv_out = T.nnet.conv.conv2d(X, self.W, border_mode=border_mode, subsample=self.subsample)
-        if self.border_mode == 'same':
-            shift_x = (self.filter_length - 1) // 2
-            conv_out = conv_out[:, :, shift_x:X.shape[2] + shift_x, :]
-
-        output = self.activation(conv_out + self.b.dimshuffle('x', 0, 'x', 'x'))
-        output = T.reshape(output, (output.shape[0], output.shape[1], output.shape[2])).dimshuffle(0, 2, 1)
+        output = conv_out + K.reshape(self.b, (1, self.nb_filter, 1, 1))
+        output = self.activation(output)
+        output = K.squeeze(output, 3)  # remove the dummy 3rd dimension
+        output = K.permute_dimensions(output, (0, 2, 1))
         return output
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "nb_filter": self.nb_filter,
-                "filter_length": self.filter_length,
-                "init": self.init.__name__,
-                "activation": self.activation.__name__,
-                "border_mode": self.border_mode,
-                "subsample_length": self.subsample_length,
-                "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
-                "b_regularizer": self.b_regularizer.get_config() if self.b_regularizer else None,
-                "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
-                "W_constraint": self.W_constraint.get_config() if self.W_constraint else None,
-                "b_constraint": self.b_constraint.get_config() if self.b_constraint else None}
+        config = {"name": self.__class__.__name__,
+                  "nb_filter": self.nb_filter,
+                  "filter_length": self.filter_length,
+                  "init": self.init.__name__,
+                  "activation": self.activation.__name__,
+                  "border_mode": self.border_mode,
+                  "subsample_length": self.subsample_length,
+                  "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
+                  "b_regularizer": self.b_regularizer.get_config() if self.b_regularizer else None,
+                  "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
+                  "W_constraint": self.W_constraint.get_config() if self.W_constraint else None,
+                  "b_constraint": self.b_constraint.get_config() if self.b_constraint else None,
+                  "input_dim": self.input_dim,
+                  "input_length": self.input_length}
+        base_config = super(Convolution1D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class Convolution2D(Layer):
-    def __init__(self, nb_filter, stack_size, nb_row, nb_col,
+    input_ndim = 4
+
+    def __init__(self, nb_filter, nb_row, nb_col,
                  init='glorot_uniform', activation='linear', weights=None,
-                 border_mode='valid', subsample=(1, 1),
+                 border_mode='valid', subsample=(1, 1), dim_ordering='th',
                  W_regularizer=None, b_regularizer=None, activity_regularizer=None,
-                 W_constraint=None, b_constraint=None):
+                 W_constraint=None, b_constraint=None, **kwargs):
 
-        if border_mode not in {'valid', 'full', 'same'}:
+        if border_mode not in {'valid', 'same'}:
             raise Exception('Invalid border mode for Convolution2D:', border_mode)
-
-        super(Convolution2D, self).__init__()
-        self.init = initializations.get(init)
-        self.activation = activations.get(activation)
-        self.subsample = subsample
-        self.border_mode = border_mode
         self.nb_filter = nb_filter
-        self.stack_size = stack_size
-
         self.nb_row = nb_row
         self.nb_col = nb_col
-
-        self.input = T.tensor4()
-        self.W_shape = (nb_filter, stack_size, nb_row, nb_col)
-        self.W = self.init(self.W_shape)
-        self.b = shared_zeros((nb_filter,))
-
-        self.params = [self.W, self.b]
-
-        self.regularizers = []
+        self.init = initializations.get(init)
+        self.activation = activations.get(activation)
+        assert border_mode in {'valid', 'same'}, 'border_mode must be in {valid, same}'
+        self.border_mode = border_mode
+        self.subsample = tuple(subsample)
+        assert dim_ordering in {'tf', 'th'}, 'dim_ordering must be in {tf, th}'
+        self.dim_ordering = dim_ordering
 
         self.W_regularizer = regularizers.get(W_regularizer)
-        if self.W_regularizer:
-            self.W_regularizer.set_param(self.W)
-            self.regularizers.append(self.W_regularizer)
-
         self.b_regularizer = regularizers.get(b_regularizer)
-        if self.b_regularizer:
-            self.b_regularizer.set_param(self.b)
-            self.regularizers.append(self.b_regularizer)
-
         self.activity_regularizer = regularizers.get(activity_regularizer)
-        if self.activity_regularizer:
-            self.activity_regularizer.set_layer(self)
-            self.regularizers.append(self.activity_regularizer)
 
         self.W_constraint = constraints.get(W_constraint)
         self.b_constraint = constraints.get(b_constraint)
         self.constraints = [self.W_constraint, self.b_constraint]
 
-        if weights is not None:
-            self.set_weights(weights)
+        self.initial_weights = weights
+        self.input = K.placeholder(ndim=4)
+        super(Convolution2D, self).__init__(**kwargs)
 
-    def get_output(self, train):
-        X = self.get_input(train)
-        border_mode = self.border_mode
-        if dnn.dnn_available() and theano.config.device[:3] == 'gpu':
-            if border_mode == 'same':
-                assert(self.subsample == (1, 1))
-                pad_x = (self.nb_row - self.subsample[0]) // 2
-                pad_y = (self.nb_col - self.subsample[1]) // 2
-                conv_out = dnn.dnn_conv(img=X,
-                                        kerns=self.W,
-                                        border_mode=(pad_x, pad_y))
-            else:
-                conv_out = dnn.dnn_conv(img=X,
-                                        kerns=self.W,
-                                        border_mode=border_mode,
-                                        subsample=self.subsample)
+    def build(self):
+        if self.dim_ordering == 'th':
+            stack_size = self.input_shape[1]
+            self.W_shape = (self.nb_filter, stack_size, self.nb_row, self.nb_col)
+        elif self.dim_ordering == 'tf':
+            stack_size = self.input_shape[3]
+            self.W_shape = (self.nb_row, self.nb_col, stack_size, self.nb_filter)
         else:
-            if border_mode == 'same':
-                border_mode = 'full'
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+        self.W = self.init(self.W_shape)
+        self.b = K.zeros((self.nb_filter,))
+        self.params = [self.W, self.b]
+        self.regularizers = []
 
-            conv_out = T.nnet.conv.conv2d(X, self.W,
-                                          border_mode=border_mode,
-                                          subsample=self.subsample)
-            if self.border_mode == 'same':
-                shift_x = (self.nb_row - 1) // 2
-                shift_y = (self.nb_col - 1) // 2
-                conv_out = conv_out[:, :, shift_x:X.shape[2] + shift_x, shift_y:X.shape[3] + shift_y]
+        if self.W_regularizer:
+            self.W_regularizer.set_param(self.W)
+            self.regularizers.append(self.W_regularizer)
 
-        return self.activation(conv_out + self.b.dimshuffle('x', 0, 'x', 'x'))
+        if self.b_regularizer:
+            self.b_regularizer.set_param(self.b)
+            self.regularizers.append(self.b_regularizer)
+
+        if self.activity_regularizer:
+            self.activity_regularizer.set_layer(self)
+            self.regularizers.append(self.activity_regularizer)
+
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.dim_ordering == 'th':
+            rows = input_shape[2]
+            cols = input_shape[3]
+        elif self.dim_ordering == 'tf':
+            rows = input_shape[1]
+            cols = input_shape[2]
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+        rows = conv_output_length(rows, self.nb_row,
+                                  self.border_mode, self.subsample[0])
+        cols = conv_output_length(cols, self.nb_col,
+                                  self.border_mode, self.subsample[1])
+
+        if self.dim_ordering == 'th':
+            return (input_shape[0], self.nb_filter, rows, cols)
+        elif self.dim_ordering == 'tf':
+            return (input_shape[0], rows, cols, self.nb_filter)
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+    def get_output(self, train=False):
+        X = self.get_input(train)
+        conv_out = K.conv2d(X, self.W, strides=self.subsample,
+                            border_mode=self.border_mode,
+                            dim_ordering=self.dim_ordering)
+
+        output = conv_out + K.reshape(self.b, (1, self.nb_filter, 1, 1))
+        output = self.activation(output)
+        return output
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "nb_filter": self.nb_filter,
-                "stack_size": self.stack_size,
-                "nb_row": self.nb_row,
-                "nb_col": self.nb_col,
-                "init": self.init.__name__,
-                "activation": self.activation.__name__,
-                "border_mode": self.border_mode,
-                "subsample": self.subsample,
-                "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
-                "b_regularizer": self.b_regularizer.get_config() if self.b_regularizer else None,
-                "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
-                "W_constraint": self.W_constraint.get_config() if self.W_constraint else None,
-                "b_constraint": self.b_constraint.get_config() if self.b_constraint else None}
+        config = {"name": self.__class__.__name__,
+                  "nb_filter": self.nb_filter,
+                  "nb_row": self.nb_row,
+                  "nb_col": self.nb_col,
+                  "init": self.init.__name__,
+                  "activation": self.activation.__name__,
+                  "border_mode": self.border_mode,
+                  "subsample": self.subsample,
+                  "dim_ordering": self.dim_ordering,
+                  "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
+                  "b_regularizer": self.b_regularizer.get_config() if self.b_regularizer else None,
+                  "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
+                  "W_constraint": self.W_constraint.get_config() if self.W_constraint else None,
+                  "b_constraint": self.b_constraint.get_config() if self.b_constraint else None}
+        base_config = super(Convolution2D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class MaxPooling1D(Layer):
-    def __init__(self, pool_length=2, stride=None, ignore_border=True):
-        super(MaxPooling1D, self).__init__()
+    input_ndim = 3
+
+    def __init__(self, pool_length=2, stride=None,
+                 border_mode='valid', **kwargs):
+        super(MaxPooling1D, self).__init__(**kwargs)
+        if stride is None:
+            stride = pool_length
         self.pool_length = pool_length
         self.stride = stride
-        if self.stride:
-            self.st = (self.stride, 1)
-        else:
-            self.st = None
+        self.st = (self.stride, 1)
+        self.input = K.placeholder(ndim=3)
+        self.pool_size = (pool_length, 1)
+        assert border_mode in {'valid', 'same'}, 'border_mode must be in {valid, same}'
+        self.border_mode = border_mode
 
-        self.input = T.tensor3()
-        self.poolsize = (pool_length, 1)
-        self.ignore_border = ignore_border
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        length = conv_output_length(input_shape[1], self.pool_length,
+                                    self.border_mode, self.stride)
+        return (input_shape[0], length, input_shape[2])
 
-    def get_output(self, train):
+    def get_output(self, train=False):
         X = self.get_input(train)
-        X = T.reshape(X, (X.shape[0], X.shape[1], X.shape[2], 1)).dimshuffle(0, 2, 1, 3)
-        output = T.signal.downsample.max_pool_2d(X, ds=self.poolsize, st=self.st, ignore_border=self.ignore_border)
-        output = output.dimshuffle(0, 2, 1, 3)
-        return T.reshape(output, (output.shape[0], output.shape[1], output.shape[2]))
+        X = K.expand_dims(X, -1)   # add dummy last dimension
+        X = K.permute_dimensions(X, (0, 2, 1, 3))
+        output = K.maxpool2d(X, pool_size=self.pool_size, strides=self.st,
+                             border_mode=self.border_mode,
+                             dim_ordering='th')
+        output = K.permute_dimensions(output, (0, 2, 1, 3))
+        return K.squeeze(output, 3)  # remove dummy last dimension
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "stride": self.stride,
-                "pool_length": self.pool_length,
-                "ignore_border": self.ignore_border}
+        config = {"name": self.__class__.__name__,
+                  "stride": self.stride,
+                  "pool_length": self.pool_length,
+                  "border_mode": self.border_mode}
+        base_config = super(MaxPooling1D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class MaxPooling2D(Layer):
-    def __init__(self, poolsize=(2, 2), stride=None, ignore_border=True):
-        super(MaxPooling2D, self).__init__()
-        self.input = T.tensor4()
-        self.poolsize = poolsize
-        self.stride = stride
-        self.ignore_border = ignore_border
+    input_ndim = 4
 
-    def get_output(self, train):
+    def __init__(self, pool_size=(2, 2), strides=None, border_mode='valid',
+                 dim_ordering='th', **kwargs):
+        super(MaxPooling2D, self).__init__(**kwargs)
+        self.input = K.placeholder(ndim=4)
+        self.pool_size = tuple(pool_size)
+        if strides is None:
+            strides = self.pool_size
+        self.strides = tuple(strides)
+        assert border_mode in {'valid', 'same'}, 'border_mode must be in {valid, same}'
+        self.border_mode = border_mode
+        assert dim_ordering in {'tf', 'th'}, 'dim_ordering must be in {tf, th}'
+        self.dim_ordering = dim_ordering
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.dim_ordering == 'th':
+            rows = input_shape[2]
+            cols = input_shape[3]
+        elif self.dim_ordering == 'tf':
+            rows = input_shape[1]
+            cols = input_shape[2]
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+        rows = conv_output_length(rows, self.pool_size[0],
+                                  self.border_mode, self.strides[0])
+        cols = conv_output_length(cols, self.pool_size[1],
+                                  self.border_mode, self.strides[1])
+
+        if self.dim_ordering == 'th':
+            return (input_shape[0], input_shape[1], rows, cols)
+        elif self.dim_ordering == 'tf':
+            return (input_shape[0], rows, cols, input_shape[3])
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+    def get_output(self, train=False):
         X = self.get_input(train)
-        output = T.signal.downsample.max_pool_2d(X, ds=self.poolsize, st=self.stride, ignore_border=self.ignore_border)
+        output = K.maxpool2d(X, pool_size=self.pool_size,
+                             strides=self.strides,
+                             border_mode=self.border_mode,
+                             dim_ordering=self.dim_ordering)
         return output
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "poolsize": self.poolsize,
-                "ignore_border": self.ignore_border,
-                "stride": self.stride}
+        config = {"name": self.__class__.__name__,
+                  "pool_size": self.pool_size,
+                  "border_mode": self.border_mode,
+                  "strides": self.strides,
+                  "dim_ordering": self.dim_ordering}
+        base_config = super(MaxPooling2D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
-class UpSample1D(Layer):
-    def __init__(self, length=2):
-        super(UpSample1D, self).__init__()
+class UpSampling1D(Layer):
+    input_ndim = 3
+
+    def __init__(self, length=2, **kwargs):
+        super(UpSampling1D, self).__init__(**kwargs)
         self.length = length
-        self.input = T.tensor3()
+        self.input = K.placeholder(ndim=3)
 
-    def get_output(self, train):
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        return (input_shape[0], self.length * input_shape[1], input_shape[2])
+
+    def get_output(self, train=False):
         X = self.get_input(train)
-        output = theano.tensor.extra_ops.repeat(X, self.length, axis=1)
+        output = K.concatenate([X] * self.length, axis=1)
         return output
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "length": self.length}
+        config = {"name": self.__class__.__name__,
+                  "length": self.length}
+        base_config = super(UpSampling1D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
-class UpSample2D(Layer):
-    def __init__(self, size=(2, 2)):
-        super(UpSample2D, self).__init__()
-        self.input = T.tensor4()
-        self.size = size
+class UpSampling2D(Layer):
+    input_ndim = 4
 
-    def get_output(self, train):
+    def __init__(self, size=(2, 2), dim_ordering='th', **kwargs):
+        super(UpSampling2D, self).__init__(**kwargs)
+        self.input = K.placeholder(ndim=4)
+        self.size = tuple(size)
+        assert dim_ordering in {'tf', 'th'}, 'dim_ordering must be in {tf, th}'
+        self.dim_ordering = dim_ordering
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.dim_ordering == 'th':
+            return (input_shape[0],
+                    input_shape[1],
+                    self.size[0] * input_shape[2],
+                    self.size[1] * input_shape[3])
+        elif self.dim_ordering == 'tf':
+            return (input_shape[0],
+                    self.size[0] * input_shape[1],
+                    self.size[1] * input_shape[2],
+                    input_shape[3])
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+    def get_output(self, train=False):
         X = self.get_input(train)
-        Y = theano.tensor.extra_ops.repeat(X, self.size[0], axis=2)
-        output = theano.tensor.extra_ops.repeat(Y, self.size[1], axis=3)
+        if self.dim_ordering == 'th':
+            output = K.concatenate([X] * self.size[0], axis=2)
+            output = K.concatenate([output] * self.size[1], axis=3)
+        elif self.dim_ordering == 'tf':
+            output = K.concatenate([X] * self.size[0], axis=1)
+            output = K.concatenate([output] * self.size[1], axis=2)
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
         return output
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "size": self.size}
+        config = {"name": self.__class__.__name__,
+                  "size": self.size}
+        base_config = super(UpSampling2D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
+
+
+class ZeroPadding1D(Layer):
+    """Zero-padding layer for 1D input (e.g. temporal sequence).
+
+    Input shape
+    -----------
+    3D tensor with shape (samples, axis_to_pad, features)
+
+    Output shape
+    ------------
+    3D tensor with shape (samples, padded_axis, features)
+
+    Arguments
+    ---------
+    padding: int
+        How many zeros to add at the beginning and end of
+        the padding dimension (axis 1).
+    """
+    input_ndim = 3
+
+    def __init__(self, padding=1, **kwargs):
+        super(ZeroPadding1D, self).__init__(**kwargs)
+        self.padding = padding
+        self.input = K.placeholder(ndim=3)
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        return (input_shape[0],
+                input_shape[1] + self.padding * 2,
+                input_shape[2])
+
+    def get_output(self, train=False):
+        X = self.get_input(train)
+        return K.temporal_padding(X, padding=self.padding)
+
+    def get_config(self):
+        config = {"name": self.__class__.__name__,
+                  "padding": self.padding}
+        base_config = super(ZeroPadding1D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class ZeroPadding2D(Layer):
-    def __init__(self, pad=(1, 1)):
-        super(ZeroPadding2D, self).__init__()
-        self.pad = pad
-        self.input = T.tensor4()
+    """Zero-padding layer for 2D input (e.g. picture).
 
-    def get_output(self, train):
+    Input shape
+    -----------
+    4D tensor with shape:
+        (samples, depth, first_axis_to_pad, second_axis_to_pad)
+
+    Output shape
+    ------------
+    4D tensor with shape:
+        (samples, depth, first_padded_axis, second_padded_axis)
+
+    Arguments
+    ---------
+    padding: tuple of int (length 2)
+        How many zeros to add at the beginning and end of
+        the 2 padding dimensions (axis 3 and 4).
+    """
+    input_ndim = 4
+
+    def __init__(self, padding=(1, 1), dim_ordering='th', **kwargs):
+        super(ZeroPadding2D, self).__init__(**kwargs)
+        self.padding = tuple(padding)
+        self.input = K.placeholder(ndim=4)
+        assert dim_ordering in {'tf', 'th'}, 'dim_ordering must be in {tf, th}'
+        self.dim_ordering = dim_ordering
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.dim_ordering == 'th':
+            return (input_shape[0],
+                    input_shape[1],
+                    input_shape[2] + 2 * self.padding[0],
+                    input_shape[3] + 2 * self.padding[1])
+        elif self.dim_ordering == 'tf':
+            return (input_shape[0],
+                    input_shape[1] + 2 * self.padding[0],
+                    input_shape[2] + 2 * self.padding[1],
+                    input_shape[3])
+        else:
+            raise Exception('Invalid dim_ordering: ' + self.dim_ordering)
+
+    def get_output(self, train=False):
         X = self.get_input(train)
-        pad = self.pad
-        in_shape = X.shape
-        out_shape = (in_shape[0], in_shape[1], in_shape[2] + 2 * pad[0], in_shape[3] + 2 * pad[1])
-        out = T.zeros(out_shape)
-        indices = (slice(None), slice(None), slice(pad[0], in_shape[2] + pad[0]), slice(pad[1], in_shape[3] + pad[1]))
-        return T.set_subtensor(out[indices], X)
+        return K.spatial_2d_padding(X, padding=self.padding,
+                                    dim_ordering=self.dim_ordering)
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "pad": self.pad}
+        config = {"name": self.__class__.__name__,
+                  "padding": self.padding}
+        base_config = super(ZeroPadding2D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/layers/embeddings.py

@@ -1,10 +1,8 @@
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
+from .. import backend as K
 
 from .. import activations, initializations, regularizers, constraints
 from ..layers.core import Layer, MaskedLayer
-from ..utils.theano_utils import sharedX
 
 from ..constraints import unitnorm
 
@@ -17,117 +15,74 @@ class Embedding(Layer):
         @input_dim: size of vocabulary (highest input integer + 1)
         @out_dim: size of dense representation
     '''
-    def __init__(self, input_dim, output_dim, init='uniform',
-                 W_regularizer=None, activity_regularizer=None, W_constraint=None,
-                 mask_zero=False, weights=None):
+    input_ndim = 2
 
-        super(Embedding, self).__init__()
-        self.init = initializations.get(init)
+    def __init__(self, input_dim, output_dim,
+                 init='uniform', input_length=None,
+                 W_regularizer=None, activity_regularizer=None,
+                 W_constraint=None,
+                 mask_zero=False,
+                 weights=None, **kwargs):
         self.input_dim = input_dim
         self.output_dim = output_dim
-
-        self.input = T.imatrix()
-        self.W = self.init((self.input_dim, self.output_dim))
+        self.init = initializations.get(init)
+        self.input_length = input_length
         self.mask_zero = mask_zero
 
-        self.params = [self.W]
-
         self.W_constraint = constraints.get(W_constraint)
         self.constraints = [self.W_constraint]
 
-        self.regularizers = []
-
         self.W_regularizer = regularizers.get(W_regularizer)
+        self.activity_regularizer = regularizers.get(activity_regularizer)
+
+        self.initial_weights = weights
+        kwargs['input_shape'] = (self.input_dim,)
+        super(Embedding, self).__init__(**kwargs)
+
+    def build(self):
+        self.input = K.placeholder(shape=(None, self.input_length),
+                                   dtype='int32')
+        self.W = self.init((self.input_dim, self.output_dim))
+        self.params = [self.W]
+        self.regularizers = []
         if self.W_regularizer:
             self.W_regularizer.set_param(self.W)
             self.regularizers.append(self.W_regularizer)
 
-        self.activity_regularizer = regularizers.get(activity_regularizer)
         if self.activity_regularizer:
             self.activity_regularizer.set_layer(self)
             self.regularizers.append(self.activity_regularizer)
 
-        if weights is not None:
-            self.set_weights(weights)
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
 
     def get_output_mask(self, train=None):
         X = self.get_input(train)
         if not self.mask_zero:
             return None
         else:
-            return T.ones_like(X) * (1 - T.eq(X, 0))
+            if K._BACKEND == "tensorflow":
+                raise Exception("Masking is Theano-only for the time being.")
+            return K.ones_like(X) * (1 - K.equal(X, 0))
+
+    @property
+    def output_shape(self):
+        return (self.input_shape[0], self.input_length, self.output_dim)
 
     def get_output(self, train=False):
         X = self.get_input(train)
-        out = self.W[X]
+        out = K.gather(self.W, X)
         return out
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
-                "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
-                "W_constraint": self.W_constraint.get_config() if self.W_constraint else None}
-
-
-class WordContextProduct(Layer):
-    '''
-        This layer turns a pair of words (a pivot word + a context word,
-        ie. a word from the same context, or a random, out-of-context word),
-        indentified by their index in a vocabulary, into two dense reprensentations
-        (word representation and context representation).
-
-        Then it returns activation(dot(pivot_embedding, context_embedding)),
-        which can be trained to encode the probability
-        of finding the context word in the context of the pivot word
-        (or reciprocally depending on your training procedure).
-
-        The layer ingests integer tensors of shape:
-        (nb_samples, 2)
-        and outputs a float tensor of shape
-        (nb_samples, 1)
-
-        The 2nd dimension encodes (pivot, context).
-        input_dim is the size of the vocabulary.
-
-        For more context, see Mikolov et al.:
-            Efficient Estimation of Word reprensentations in Vector Space
-            http://arxiv.org/pdf/1301.3781v3.pdf
-    '''
-    def __init__(self, input_dim, proj_dim=128,
-                 init='uniform', activation='sigmoid', weights=None):
-
-        super(WordContextProduct, self).__init__()
-        self.input_dim = input_dim
-        self.proj_dim = proj_dim
-        self.init = initializations.get(init)
-        self.activation = activations.get(activation)
-
-        self.input = T.imatrix()
-        # two different embeddings for pivot word and its context
-        # because p(w|c) != p(c|w)
-        self.W_w = self.init((input_dim, proj_dim))
-        self.W_c = self.init((input_dim, proj_dim))
-
-        self.params = [self.W_w, self.W_c]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        w = self.W_w[X[:, 0]]  # nb_samples, proj_dim
-        c = self.W_c[X[:, 1]]  # nb_samples, proj_dim
-
-        dot = T.sum(w * c, axis=1)
-        dot = theano.tensor.reshape(dot, (X.shape[0], 1))
-        return self.activation(dot)
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "proj_dim": self.proj_dim,
-                "init": self.init.__name__,
-                "activation": self.activation.__name__}
+        config = {"name": self.__class__.__name__,
+                  "input_dim": self.input_dim,
+                  "output_dim": self.output_dim,
+                  "init": self.init.__name__,
+                  "input_length": self.input_length,
+                  "mask_zero": self.mask_zero,
+                  "activity_regularizer": self.activity_regularizer.get_config() if self.activity_regularizer else None,
+                  "W_regularizer": self.W_regularizer.get_config() if self.W_regularizer else None,
+                  "W_constraint": self.W_constraint.get_config() if self.W_constraint else None}
+        base_config = super(Embedding, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/layers/noise.py

@@ -1,31 +1,30 @@
 from __future__ import absolute_import
-import numpy as np
 from .core import MaskedLayer
-import theano
-import theano.tensor as T
-from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams
+from .. import backend as K
 
 
 class GaussianNoise(MaskedLayer):
     '''
         Corruption process with GaussianNoise
     '''
-    def __init__(self, sigma):
-        super(GaussianNoise, self).__init__()
+    def __init__(self, sigma, **kwargs):
+        super(GaussianNoise, self).__init__(**kwargs)
         self.sigma = sigma
-        self.srng = RandomStreams(seed=np.random.randint(10e6))
 
     def get_output(self, train=False):
         X = self.get_input(train)
         if not train or self.sigma == 0:
             return X
         else:
-            return X + self.srng.normal(size=X.shape, avg=0.0, std=self.sigma,
-                                        dtype=theano.config.floatX)
+            return X + K.random_normal(shape=K.shape(X),
+                                       mean=0.,
+                                       std=self.sigma)
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "sigma": self.sigma}
+        config = {"name": self.__class__.__name__,
+                  "sigma": self.sigma}
+        base_config = super(GaussianNoise, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class GaussianDropout(MaskedLayer):
@@ -36,18 +35,21 @@ class GaussianDropout(MaskedLayer):
             Srivastava, Hinton, et al. 2014
             http://www.cs.toronto.edu/~rsalakhu/papers/srivastava14a.pdf
     '''
-    def __init__(self, p):
-        super(GaussianDropout, self).__init__()
+    def __init__(self, p, **kwargs):
+        super(GaussianDropout, self).__init__(**kwargs)
         self.p = p
-        self.srng = RandomStreams(seed=np.random.randint(10e6))
 
     def get_output(self, train):
         X = self.get_input(train)
         if train:
-            # self.p refers to drop probability rather than retain probability (as in paper) to match Dropout layer syntax
-            X *= self.srng.normal(size=X.shape, avg=1.0, std=T.sqrt(self.p / (1.0 - self.p)), dtype=theano.config.floatX)
+            # self.p refers to drop probability rather than
+            # retain probability (as in paper), for consistency
+            X *= K.random_normal(shape=K.shape(X), mean=1.0,
+                                 std=self.p / (1.0 - self.p))
         return X
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "p": self.p}
+        config = {"name": self.__class__.__name__,
+                  "p": self.p}
+        base_config = super(GaussianDropout, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/layers/normalization.py

@@ -1,66 +1,85 @@
 from ..layers.core import Layer
-from ..utils.theano_utils import shared_zeros, shared_ones, ndim_tensor
 from .. import initializations
-
-import theano.tensor as T
+from .. import backend as K
 
 
 class BatchNormalization(Layer):
     '''
         Reference:
-            Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
+            Batch Normalization: Accelerating Deep Network Training
+            by Reducing Internal Covariate Shift
                 http://arxiv.org/pdf/1502.03167v3.pdf
 
             mode: 0 -> featurewise normalization
-                  1 -> samplewise normalization (may sometimes outperform featurewise mode)
+                  1 -> samplewise normalization
+                       (may sometimes outperform featurewise mode)
 
-            momentum: momentum term in the computation of a running estimate of the mean and std of the data
+            momentum: momentum term in the computation
+            of a running estimate of the mean and std of the data
     '''
-    def __init__(self, input_shape, epsilon=1e-6, mode=0, momentum=0.9, weights=None):
-        super(BatchNormalization, self).__init__()
+    def __init__(self, epsilon=1e-6, mode=0, momentum=0.9,
+                 weights=None, **kwargs):
         self.init = initializations.get("uniform")
-        self.input_shape = input_shape
         self.epsilon = epsilon
         self.mode = mode
         self.momentum = momentum
-        self.input = ndim_tensor(len(self.input_shape) + 1)
+        self.initial_weights = weights
+        super(BatchNormalization, self).__init__(**kwargs)
 
-        self.gamma = self.init((self.input_shape))
-        self.beta = shared_zeros(self.input_shape)
+    def build(self):
+        input_shape = self.input_shape  # starts with samples axis
+        input_shape = input_shape[1:]
+
+        self.gamma = self.init((input_shape))
+        self.beta = K.zeros(input_shape)
 
         self.params = [self.gamma, self.beta]
-        if weights is not None:
-            self.set_weights(weights)
+        self.running_mean = K.zeros(input_shape)
+        self.running_std = K.ones((input_shape))
 
-    def init_updates(self):
-        self.running_mean = shared_zeros(self.input_shape)
-        self.running_std = shared_ones((self.input_shape))
+        # initialize self.updates: batch mean/std computation
         X = self.get_input(train=True)
-        m = X.mean(axis=0)
-        std = T.mean((X - m) ** 2 + self.epsilon, axis=0) ** 0.5
+        m = K.mean(X, axis=0)
+        std = K.mean(K.square(X - m) + self.epsilon, axis=0)
+        std = K.sqrt(std)
         mean_update = self.momentum * self.running_mean + (1-self.momentum) * m
         std_update = self.momentum * self.running_std + (1-self.momentum) * std
-        self.updates = [(self.running_mean, mean_update), (self.running_std, std_update)]
+        self.updates = [(self.running_mean, mean_update),
+                        (self.running_std, std_update)]
+
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
+
+    def get_weights(self):
+        super_weights = super(BatchNormalization, self).get_weights()
+        return super_weights + [K.get_value(self.running_mean),
+                                K.get_value(self.running_std)]
+
+    def set_weights(self, weights):
+        K.set_value(self.running_mean, weights[-2])
+        K.set_value(self.running_std, weights[-1])
+        super(BatchNormalization, self).set_weights(weights[:-2])
 
     def get_output(self, train):
         X = self.get_input(train)
-
         if self.mode == 0:
-            X_normed = (X - self.running_mean) / (self.running_std + self.epsilon)
-
+            X_normed = ((X - self.running_mean) /
+                        (self.running_std + self.epsilon))
         elif self.mode == 1:
-            m = X.mean(axis=-1, keepdims=True)
-            std = X.std(axis=-1, keepdims=True)
+            m = K.mean(X, axis=-1, keepdims=True)
+            std = K.std(X, axis=-1, keepdims=True)
             X_normed = (X - m) / (std + self.epsilon)
-
         out = self.gamma * X_normed + self.beta
         return out
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_shape": self.input_shape,
-                "epsilon": self.epsilon,
-                "mode": self.mode}
+        config = {"name": self.__class__.__name__,
+                  "epsilon": self.epsilon,
+                  "mode": self.mode,
+                  "momentum": self.momentum}
+        base_config = super(BatchNormalization, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class LRN2D(Layer):
@@ -69,10 +88,10 @@ class LRN2D(Layer):
     License at: https://github.com/lisa-lab/pylearn2/blob/master/LICENSE.txt
     """
 
-    def __init__(self, alpha=1e-4, k=2, beta=0.75, n=5):
+    def __init__(self, alpha=1e-4, k=2, beta=0.75, n=5, **kwargs):
         if n % 2 == 0:
             raise NotImplementedError("LRN2D only works with odd n. n provided: " + str(n))
-        super(LRN2D, self).__init__()
+        super(LRN2D, self).__init__(**kwargs)
         self.alpha = alpha
         self.k = k
         self.beta = beta
@@ -80,20 +99,25 @@ class LRN2D(Layer):
 
     def get_output(self, train):
         X = self.get_input(train)
-        b, ch, r, c = X.shape
+        b, ch, r, c = K.shape(X)
         half_n = self.n // 2
-        input_sqr = T.sqr(X)
-        extra_channels = T.alloc(0., b, ch + 2*half_n, r, c)
-        input_sqr = T.set_subtensor(extra_channels[:, half_n:half_n+ch, :, :], input_sqr)
+        input_sqr = K.square(X)
+        extra_channels = K.zeros((b, ch + 2 * half_n, r, c))
+        input_sqr = K.concatenate([extra_channels[:, :half_n, :, :],
+                                   input_sqr,
+                                   extra_channels[:, half_n + ch:, :, :]],
+                                  axis=1)
         scale = self.k
         for i in range(self.n):
-            scale += self.alpha * input_sqr[:, i:i+ch, :, :]
+            scale += self.alpha * input_sqr[:, i:i + ch, :, :]
         scale = scale ** self.beta
         return X / scale
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "alpha": self.alpha,
-                "k": self.k,
-                "beta": self.beta,
-                "n": self.n}
+        config = {"name": self.__class__.__name__,
+                  "alpha": self.alpha,
+                  "k": self.k,
+                  "beta": self.beta,
+                  "n": self.n}
+        base_config = super(LRN2D, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/layers/ntm.py

@@ -0,0 +1,326 @@
+import numpy as np
+from scipy.linalg import circulant
+
+from .. import backend as K
+import theano
+import theano.tensor as T
+floatX = theano.config.floatX
+
+from keras.layers.recurrent import Recurrent, GRU, LSTM
+from keras.utils.theano_utils import shared_zeros, alloc_zeros_matrix, shared_scalar
+tol = 1e-4
+
+
+def _update_controller(self, inp, h_tm1, M, mask):
+    """ Update inner RNN controler
+    We have to update the inner RNN inside the Neural Turing Machine, this
+    is an almost literal copy of keras.layers.recurrent.GRU and 
+    keras.layers.recurrent.LSTM see these clases for further details.
+    """
+    x = T.concatenate([inp, M], axis=-1)
+    # get inputs
+    if self.inner_rnn == 'gru':
+        x_z = T.dot(x, self.rnn.W_z) + self.rnn.b_z
+        x_r = T.dot(x, self.rnn.W_r) + self.rnn.b_r
+        x_h = T.dot(x, self.rnn.W_h) + self.rnn.b_h
+
+    elif self.inner_rnn == 'lstm':
+        xi = T.dot(x, self.rnn.W_i) + self.rnn.b_i
+        xf = T.dot(x, self.rnn.W_f) + self.rnn.b_f
+        xc = T.dot(x, self.rnn.W_c) + self.rnn.b_c
+        xo = T.dot(x, self.rnn.W_o) + self.rnn.b_o
+
+    elif self.inner_rnn == 'simple':
+        x = T.dot(x, self.rnn.W) + self.rnn.b
+
+    # update state
+    if self.inner_rnn == 'gru':
+        h = self.rnn._step(x_z, x_r, x_h, 1., h_tm1[0],
+                           self.rnn.U_z,
+                           self.rnn.U_r,
+                           self.rnn.U_h)
+        h = mask[:, None] * h + (1-mask[:, None])*h_tm1[0]
+        h = (h, )
+
+    elif self.inner_rnn == 'lstm':
+        h = self.rnn._step(xi, xf, xo, xc, 1.,
+                           h_tm1[1], h_tm1[0],
+                           self.rnn.U_i, self.rnn.U_f,
+                           self.rnn.U_o, self.rnn.U_c)
+        h = h[::-1]
+        h = tuple([mask[:, None]*h[i] +
+                   (1-mask[:, None])*h_tm1[i] for i in range(len(h))])
+
+    elif self.inner_rnn == 'simple':
+        h = self.rnn._step(x, 1, h_tm1[0], self.rnn.U)
+        h = mask[:, None] * h + (1-mask[:, None])*h_tm1[0]
+        h = (h, )
+
+    return h
+
+
+def _circulant(leng, n_shifts):
+    """ Generate circulant copies of a vector.
+    This will generate a tensor with `n_shifts` of rotated versions the
+    identity matrix. When this tensor is multiplied by a vector
+    the result are `n_shifts` shifted versions of that vector. Since
+    everything is done with inner products, this operation is differentiable.
+
+    Paramters:
+    ----------
+    leng: int > 0, number of memory locations
+    n_shifts: int > 0, number of allowed shifts (if 1, no shift)
+
+    Returns:
+    --------
+    shift operation, a tensor with dimensions (n_shifts, leng, leng)
+    """
+    eye = np.eye(leng)
+    shifts = range(n_shifts//2, -n_shifts//2, -1)
+    C = np.asarray([np.roll(eye, s, axis=1) for s in shifts])
+    return theano.shared(C.astype(theano.config.floatX))
+
+
+def _renorm(x):
+    return x / (x.sum(axis=1, keepdims=True))
+
+
+def _softmax(x):
+    wt = x.flatten(ndim=2)
+    w = T.nnet.softmax(wt)
+    return w.reshape(x.shape)  # T.clip(s, 0, 1)
+
+
+def _cosine_distance(M, k):
+    dot = (M * k[:, None, :]).sum(axis=-1)
+    nM = T.sqrt((M**2).sum(axis=-1))
+    nk = T.sqrt((k**2).sum(axis=-1, keepdims=True))
+    return dot / (nM * nk)
+
+
+class NeuralTuringMachine(Recurrent):
+    """ Neural Turing Machines
+    
+    Parameters:
+    -----------
+    shift_range: int, number of available shifts, ex. if 3, avilable shifts are
+                 (-1, 0, 1)
+    n_slots: number of memory locations
+    m_length: memory length at each location
+    inner_rnn: str, supported values are 'gru' and 'lstm'
+    output_dim: hidden state size (RNN controller output_dim)
+
+    Known issues and TODO:
+    ----------------------
+    Theano may complain when n_slots == 1.
+    Add multiple reading and writing heads.
+
+    """
+    def __init__(self, output_dim, n_slots, m_length, shift_range=3,
+                 inner_rnn='gru', truncate_gradient=-1, return_sequences=False,
+                 init='glorot_uniform', inner_init='orthogonal',
+                 input_dim=None, input_length=None, **kwargs):
+        if K._BACKEND != 'theano':
+            raise Exception('NeuralTuringMachine is only available for Theano for the time being. ' +
+                            'It will be adapted to TensorFlow soon.')
+        self.output_dim = output_dim
+        self.n_slots = n_slots
+        self.m_length = m_length
+        self.shift_range = shift_range
+        self.init = init
+        self.inner_init = inner_init
+        self.inner_rnn = inner_rnn
+        self.return_sequences = return_sequences
+        self.truncate_gradient = truncate_gradient
+
+        self.input_dim = input_dim
+        self.input_length = input_length
+        if self.input_dim:
+            kwargs['input_shape'] = (self.input_length, self.input_dim)
+        super(NeuralTuringMachine, self).__init__(**kwargs)
+
+    def build(self):
+        input_leng, input_dim = self.input_shape[1:]
+        self.input = T.tensor3()
+
+        if self.inner_rnn == 'gru':
+            self.rnn = GRU(
+                input_dim=input_dim+self.m_length,
+                input_length=input_leng,
+                output_dim=self.output_dim, init=self.init,
+                inner_init=self.inner_init)
+        elif self.inner_rnn == 'lstm':
+            self.rnn = LSTM(
+                input_dim=input_dim+self.m_length,
+                input_length=input_leng,
+                output_dim=self.output_dim, init=self.init,
+                inner_init=self.inner_init)
+        else:
+            raise ValueError('this inner_rnn is not implemented yet.')
+
+        self.rnn.build()
+
+        # initial memory, state, read and write vecotrs
+        self.M = theano.shared((.001*np.ones((1,)).astype(floatX)))
+        self.init_h = shared_zeros((self.output_dim))
+        self.init_wr = self.rnn.init((self.n_slots,))
+        self.init_ww = self.rnn.init((self.n_slots,))
+
+        # write
+        self.W_e = self.rnn.init((self.output_dim, self.m_length))  # erase
+        self.b_e = shared_zeros((self.m_length))
+        self.W_a = self.rnn.init((self.output_dim, self.m_length))  # add
+        self.b_a = shared_zeros((self.m_length))
+
+        # get_w  parameters for reading operation
+        self.W_k_read = self.rnn.init((self.output_dim, self.m_length))
+        self.b_k_read = self.rnn.init((self.m_length, ))
+        self.W_c_read = self.rnn.init((self.output_dim, 3))  # 3 = beta, g, gamma see eq. 5, 7, 9 in Graves et. al 2014
+        self.b_c_read = shared_zeros((3))
+        self.W_s_read = self.rnn.init((self.output_dim, self.shift_range))
+        self.b_s_read = shared_zeros((self.shift_range)) 
+
+        # get_w  parameters for writing operation
+        self.W_k_write = self.rnn.init((self.output_dim, self.m_length))
+        self.b_k_write = self.rnn.init((self.m_length, ))
+        self.W_c_write = self.rnn.init((self.output_dim, 3))  # 3 = beta, g, gamma see eq. 5, 7, 9
+        self.b_c_write = shared_zeros((3))
+        self.W_s_write = self.rnn.init((self.output_dim, self.shift_range))
+        self.b_s_write = shared_zeros((self.shift_range))
+
+        self.C = _circulant(self.n_slots, self.shift_range)
+
+        self.params = self.rnn.params + [
+            self.W_e, self.b_e,
+            self.W_a, self.b_a,
+            self.W_k_read, self.b_k_read,
+            self.W_c_read, self.b_c_read,
+            self.W_s_read, self.b_s_read,
+            self.W_k_write, self.b_k_write,
+            self.W_s_write, self.b_s_write,
+            self.W_c_write, self.b_c_write,
+            self.M,
+            self.init_h, self.init_wr, self.init_ww]
+
+        if self.inner_rnn == 'lstm':
+            self.init_c = shared_zeros((self.output_dim))
+            self.params = self.params + [self.init_c, ]
+
+    def _read(self, w, M):
+        return (w[:, :, None]*M).sum(axis=1)
+
+    def _write(self, w, e, a, M, mask):
+        Mtilda = M * (1 - w[:, :, None]*e[:, None, :])
+        Mout = Mtilda + w[:, :, None]*a[:, None, :]
+        return mask[:, None, None]*Mout + (1-mask[:, None, None])*M
+
+    def _get_content_w(self, beta, k, M):
+        num = beta[:, None] * _cosine_distance(M, k)
+        return _softmax(num)
+
+    def _get_location_w(self, g, s, C, gamma, wc, w_tm1, mask):
+        wg = g[:, None] * wc + (1-g[:, None])*w_tm1
+        Cs = (C[None, :, :, :] * wg[:, None, None, :]).sum(axis=3)
+        wtilda = (Cs * s[:, :, None]).sum(axis=1)
+        wout = _renorm(wtilda ** gamma[:, None])
+        return mask[:, None] * wout + (1-mask[:, None])*w_tm1
+
+    def _get_controller_output(self, h, W_k, b_k, W_c, b_c, W_s, b_s):
+        k = T.tanh(T.dot(h, W_k) + b_k)  # + 1e-6
+        c = T.dot(h, W_c) + b_c
+        beta = T.nnet.relu(c[:, 0]) + 1e-6
+        g = T.nnet.sigmoid(c[:, 1])
+        gamma = T.nnet.relu(c[:, 2]) + 1
+        s = T.nnet.softmax(T.dot(h, W_s) + b_s)
+        return k, beta, g, gamma, s
+
+    def _get_initial_states(self, batch_size):
+        init_M = self.M.dimshuffle(0, 'x', 'x').repeat(
+           batch_size, axis=0).repeat(self.n_slots, axis=1).repeat(
+               self.m_length, axis=2)
+
+        init_h = self.init_h.dimshuffle(('x', 0)).repeat(batch_size, axis=0)
+        init_wr = self.init_wr.dimshuffle(('x', 0)).repeat(batch_size, axis=0)
+        init_ww = self.init_ww.dimshuffle(('x', 0)).repeat(batch_size, axis=0)
+        if self.inner_rnn == 'lstm':
+            init_c = self.init_c.dimshuffle(('x', 0)).repeat(batch_size, axis=0)
+            return init_M, T.nnet.softmax(init_wr), T.nnet.softmax(init_ww), init_h, init_c
+        else:
+            return init_M, T.nnet.softmax(init_wr), T.nnet.softmax(init_ww), init_h
+
+    def _step(self, x, mask, M_tm1, wr_tm1, ww_tm1, *args):
+        # read
+        if self.inner_rnn == 'lstm':
+            h_tm1 = args[0:2][::-1]  # (cell_tm1, h_tm1)
+        else:
+            h_tm1 = args[0:1]  # (h_tm1, )
+        k_read, beta_read, g_read, gamma_read, s_read = self._get_controller_output(
+            h_tm1[-1], self.W_k_read, self.b_k_read, self.W_c_read, self.b_c_read,
+            self.W_s_read, self.b_s_read)
+        wc_read = self._get_content_w(beta_read, k_read, M_tm1)
+        wr_t = self._get_location_w(g_read, s_read, self.C, gamma_read,
+                                    wc_read, wr_tm1, mask)
+        M_read = self._read(wr_t, M_tm1)
+
+        # update controller
+        h_t = _update_controller(self, x, h_tm1, M_read, mask)
+
+        # write
+        k_write, beta_write, g_write, gamma_write, s_write = self._get_controller_output(
+            h_t[-1], self.W_k_write, self.b_k_write, self.W_c_write,
+            self.b_c_write, self.W_s_write, self.b_s_write)
+        wc_write = self._get_content_w(beta_write, k_write, M_tm1)
+        ww_t = self._get_location_w(g_write, s_write, self.C, gamma_write,
+                                    wc_write, ww_tm1, mask)
+        e = T.nnet.sigmoid(T.dot(h_t[-1], self.W_e) + self.b_e)
+        a = T.tanh(T.dot(h_t[-1], self.W_a) + self.b_a)
+        M_t = self._write(ww_t, e, a, M_tm1, mask)
+
+        return (M_t, wr_t, ww_t) + h_t
+
+    def get_output(self, train=False):
+        outputs = self.get_full_output(train)
+
+        if self.return_sequences:
+            return outputs[-1]
+        else:
+            return outputs[-1][:, -1]
+
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.return_sequences:
+            return input_shape[0], input_shape[1], self.output_dim
+        else:
+            return input_shape[0], self.output_dim
+
+    def get_full_output(self, train=False):
+        """
+        This method is for research and visualization purposes. Use it as:
+        X = model.get_input()  # full model
+        Y = ntm.get_output()    # this layer
+        F = theano.function([X], Y, allow_input_downcast=True)
+        [memory, read_address, write_address, rnn_state] = F(x)
+
+        if inner_rnn == "lstm" use it as:
+        [memory, read_address, write_address, rnn_cell, rnn_state] = F(x)
+
+        """
+        X = self.get_input(train)
+        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)[:, :, 0]
+        X = X.dimshuffle((1, 0, 2))
+
+        init_states = self._get_initial_states(X.shape[1])
+        outputs, updates = theano.scan(self._step,
+                                       sequences=[X, padded_mask],
+                                       outputs_info=init_states,
+                                       non_sequences=self.params,
+                                       truncate_gradient=self.truncate_gradient)
+
+        out = [outputs[0].dimshuffle((1, 0, 2, 3)),
+               outputs[1].dimshuffle(1, 0, 2),
+               outputs[2].dimshuffle((1, 0, 2)),
+               outputs[3].dimshuffle((1, 0, 2))]
+        if self.inner_rnn == 'lstm':
+            out + [outputs[4].dimshuffle((1, 0, 2))]
+        return out

keras/layers/recurrent.py

@@ -1,316 +1,269 @@
 # -*- coding: utf-8 -*-
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
 import numpy as np
 
+from .. import backend as K
 from .. import activations, initializations
-from ..utils.theano_utils import shared_scalar, shared_zeros, alloc_zeros_matrix
-from ..layers.core import Layer, MaskedLayer
-from six.moves import range
+from ..layers.core import MaskedLayer
 
 
 class Recurrent(MaskedLayer):
-    def get_output_mask(self, train=None):
+    input_ndim = 3
+
+    def __init__(self, weights=None,
+                 return_sequences=False, go_backwards=False, stateful=False,
+                 input_dim=None, input_length=None, **kwargs):
+        self.return_sequences = return_sequences
+        self.initial_weights = weights
+        self.go_backwards = go_backwards
+        self.stateful = stateful
+
+        self.input_dim = input_dim
+        self.input_length = input_length
+        if self.input_dim:
+            kwargs['input_shape'] = (self.input_length, self.input_dim)
+        super(Recurrent, self).__init__(**kwargs)
+
+    def get_output_mask(self, train=False):
         if self.return_sequences:
             return super(Recurrent, self).get_output_mask(train)
         else:
             return None
 
-    def get_padded_shuffled_mask(self, train, X, pad=0):
-        mask = self.get_input_mask(train)
-        if mask is None:
-            mask = T.ones_like(X.sum(axis=-1))  # is there a better way to do this without a sum?
+    @property
+    def output_shape(self):
+        input_shape = self.input_shape
+        if self.return_sequences:
+            return (input_shape[0], input_shape[1], self.output_dim)
+        else:
+            return (input_shape[0], self.output_dim)
 
-        # mask is (nb_samples, time)
-        mask = T.shape_padright(mask)  # (nb_samples, time, 1)
-        mask = T.addbroadcast(mask, -1)  # (time, nb_samples, 1) matrix.
-        mask = mask.dimshuffle(1, 0, 2)  # (time, nb_samples, 1)
+    def step(self, x, states):
+        raise NotImplementedError
 
-        if pad > 0:
-            # left-pad in time with 0
-            padding = alloc_zeros_matrix(pad, mask.shape[1], 1)
-            mask = T.concatenate([padding, mask], axis=0)
-        return mask.astype('int8')
+    def get_output(self, train=False):
+        # input shape: (nb_samples, time (padded with zeros), input_dim)
+        X = self.get_input(train)
+        assert K.ndim(X) == 3
+        if K._BACKEND == 'tensorflow':
+            if not self.input_shape[1]:
+                raise Exception('When using TensorFlow, you should define ' +
+                                'explicitely the number of timesteps of ' +
+                                'your sequences. Make sure the first layer ' +
+                                'has an "input_shape" argument with a defined ' +
+                                'first dimension.')
+
+        mask = self.get_output_mask(train)
+        if mask:
+            # apply mask
+            X *= K.expand_dims(mask)
+            masking = True
+        else:
+            masking = False
+
+        if self.stateful:
+            initial_states = self.states
+        else:
+            # build an all-zero tensor of shape (samples, output_dim)
+            initial_state = K.zeros_like(X)  # (samples, timesteps, input_dim)
+            initial_state = K.sum(initial_state, axis=1)  # (samples, input_dim)
+            reducer = K.zeros((self.input_dim, self.output_dim))
+            initial_state = K.dot(initial_state, reducer)  # (samples, output_dim)
+            initial_states = [initial_state for _ in range(len(self.states))]
+
+        last_output, outputs, states = K.rnn(self.step, X, initial_states,
+                                             go_backwards=self.go_backwards,
+                                             masking=masking)
+        if self.stateful:
+            for i in range(len(states)):
+                K.set_value(self.states[i], states[i])
+
+        if self.return_sequences:
+            return outputs
+        else:
+            return last_output
+
+    def get_config(self):
+        config = {"name": self.__class__.__name__,
+                  "return_sequences": self.return_sequences,
+                  "input_dim": self.input_dim,
+                  "input_length": self.input_length,
+                  "go_backwards": self.go_backwards,
+                  "stateful": self.stateful}
+        base_config = super(Recurrent, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class SimpleRNN(Recurrent):
     '''
-        Fully connected RNN where output is to fed back to input.
-
-        Not a particularly useful model,
-        included for demonstration purposes
-        (demonstrates how to use theano.scan to build a basic RNN).
-    '''
-    def __init__(self, input_dim, output_dim,
-                 init='glorot_uniform', inner_init='orthogonal', activation='sigmoid', weights=None,
-                 truncate_gradient=-1, return_sequences=False):
-
-        super(SimpleRNN, self).__init__()
-        self.init = initializations.get(init)
-        self.inner_init = initializations.get(inner_init)
-        self.input_dim = input_dim
-        self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.activation = activations.get(activation)
-        self.return_sequences = return_sequences
-        self.input = T.tensor3()
-
-        self.W = self.init((self.input_dim, self.output_dim))
-        self.U = self.inner_init((self.output_dim, self.output_dim))
-        self.b = shared_zeros((self.output_dim))
-        self.params = [self.W, self.U, self.b]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def _step(self, x_t, mask_tm1, h_tm1, u):
-        '''
-            Variable names follow the conventions from:
-            http://deeplearning.net/software/theano/library/scan.html
-
-        '''
-        return self.activation(x_t + mask_tm1 * T.dot(h_tm1, u))
-
-    def get_output(self, train=False):
-        X = self.get_input(train)  # shape: (nb_samples, time (padded with zeros), input_dim)
-        # new shape: (time, nb_samples, input_dim) -> because theano.scan iterates over main dimension
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
-        x = T.dot(X, self.W) + self.b
-
-        # scan = theano symbolic loop.
-        # See: http://deeplearning.net/software/theano/library/scan.html
-        # Iterate over the first dimension of the x array (=time).
-        outputs, updates = theano.scan(
-            self._step,  # this will be called with arguments (sequences[i], outputs[i-1], non_sequences[i])
-            sequences=[x, dict(input=padded_mask, taps=[-1])],  # tensors to iterate over, inputs to _step
-            # initialization of the output. Input to _step with default tap=-1.
-            outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-            non_sequences=self.U,  # static inputs to _step
-            truncate_gradient=self.truncate_gradient)
-
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
-
-
-class SimpleDeepRNN(Recurrent):
-    '''
-        Fully connected RNN where the output of multiple timesteps
-        (up to "depth" steps in the past) is fed back to the input:
-
-        output = activation( W.x_t + b + inner_activation(U_1.h_tm1) + inner_activation(U_2.h_tm2) + ... )
-
-        This demonstrates how to build RNNs with arbitrary lookback.
-        Also (probably) not a super useful model.
-    '''
-    def __init__(self, input_dim, output_dim, depth=3,
-                 init='glorot_uniform', inner_init='orthogonal',
-                 activation='sigmoid', inner_activation='hard_sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(SimpleDeepRNN, self).__init__()
-        self.init = initializations.get(init)
-        self.inner_init = initializations.get(inner_init)
-        self.input_dim = input_dim
-        self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.activation = activations.get(activation)
-        self.inner_activation = activations.get(inner_activation)
-        self.depth = depth
-        self.return_sequences = return_sequences
-        self.input = T.tensor3()
-
-        self.W = self.init((self.input_dim, self.output_dim))
-        self.Us = [self.inner_init((self.output_dim, self.output_dim)) for _ in range(self.depth)]
-        self.b = shared_zeros((self.output_dim))
-        self.params = [self.W] + self.Us + [self.b]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def _step(self, x_t, *args):
-        o = x_t
-        for i in range(self.depth):
-            mask_tmi = args[i]
-            h_tmi = args[i + self.depth]
-            U_tmi = args[i + 2*self.depth]
-            o += mask_tmi*self.inner_activation(T.dot(h_tmi, U_tmi))
-        return self.activation(o)
-
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=self.depth)
-        X = X.dimshuffle((1, 0, 2))
-
-        x = T.dot(X, self.W) + self.b
-
-        if self.depth == 1:
-            initial = T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1)
-        else:
-            initial = T.unbroadcast(T.unbroadcast(alloc_zeros_matrix(self.depth, X.shape[1], self.output_dim), 0), 2)
-
-        outputs, updates = theano.scan(
-            self._step,
-            sequences=[x, dict(
-                input=padded_mask,
-                taps=[(-i) for i in range(self.depth)]
-            )],
-            outputs_info=[dict(
-                initial=initial,
-                taps=[(-i-1) for i in range(self.depth)]
-            )],
-            non_sequences=self.Us,
-            truncate_gradient=self.truncate_gradient
-        )
-
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "depth": self.depth,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
-
-
-class GRU(Recurrent):
-    '''
-        Gated Recurrent Unit - Cho et al. 2014
-
-        Acts as a spatiotemporal projection,
-        turning a sequence of vectors into a single vector.
-
-        Eats inputs with shape:
-        (nb_samples, max_sample_length (samples shorter than this are padded with zeros at the end), input_dim)
-
+        Fully-connected RNN where the output is to fed back to input.
+        Takes inputs with shape:
+        (nb_samples, max_sample_length, input_dim)
+        (samples shorter than `max_sample_length`
+         are padded with zeros at the end)
         and returns outputs with shape:
         if not return_sequences:
             (nb_samples, output_dim)
         if return_sequences:
             (nb_samples, max_sample_length, output_dim)
+    '''
+    def __init__(self, output_dim,
+                 init='glorot_uniform', inner_init='orthogonal',
+                 activation='sigmoid', **kwargs):
+        self.output_dim = output_dim
+        self.init = initializations.get(init)
+        self.inner_init = initializations.get(inner_init)
+        self.activation = activations.get(activation)
+        super(SimpleRNN, self).__init__(**kwargs)
 
+    def build(self):
+        input_shape = self.input_shape
+        if self.stateful:
+            if not input_shape[0]:
+                raise Exception('If a RNN is stateful, a complete ' +
+                                'input_shape must be provided ' +
+                                '(including batch size).')
+            self.states = [K.zeros(input_shape[0], self.output_dim)]
+        else:
+            # initial states: all-zero tensor of shape (output_dim)
+            self.states = [None]
+        input_dim = input_shape[2]
+        self.input_dim = input_dim
+
+        self.W = self.init((input_dim, self.output_dim))
+        self.U = self.inner_init((self.output_dim, self.output_dim))
+        self.b = K.zeros((self.output_dim))
+        self.params = [self.W, self.U, self.b]
+
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
+
+    def step(self, x, states):
+        # states only contains the previous output.
+        assert len(states) == 1
+        prev_output = states[0]
+        h = K.dot(x, self.W) + self.b
+        output = self.activation(h * K.dot(prev_output, self.U))
+        return output, [output]
+
+    def get_config(self):
+        config = {"output_dim": self.output_dim,
+                  "init": self.init.__name__,
+                  "inner_init": self.inner_init.__name__,
+                  "activation": self.activation.__name__}
+        base_config = super(SimpleRNN, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
+
+
+class GRU(Recurrent):
+    '''
+        Gated Recurrent Unit - Cho et al. 2014
+        Acts as a spatiotemporal projection,
+        turning a sequence of vectors into a single vector.
+        Takes inputs with shape:
+        (nb_samples, max_sample_length, input_dim)
+        (samples shorter than `max_sample_length`
+         are padded with zeros at the end)
+        and returns outputs with shape:
+        if not return_sequences:
+            (nb_samples, output_dim)
+        if return_sequences:
+            (nb_samples, max_sample_length, output_dim)
         References:
-            On the Properties of Neural Machine Translation: Encoder–Decoder Approaches
+            On the Properties of Neural Machine Translation:
+            Encoder–Decoder Approaches
                 http://www.aclweb.org/anthology/W14-4012
-            Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
+            Empirical Evaluation of Gated Recurrent Neural Networks
+            on Sequence Modeling
                 http://arxiv.org/pdf/1412.3555v1.pdf
     '''
-    def __init__(self, input_dim, output_dim=128,
+    def __init__(self, output_dim,
                  init='glorot_uniform', inner_init='orthogonal',
                  activation='sigmoid', inner_activation='hard_sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(GRU, self).__init__()
-        self.input_dim = input_dim
+                 **kwargs):
         self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.return_sequences = return_sequences
-
         self.init = initializations.get(init)
         self.inner_init = initializations.get(inner_init)
         self.activation = activations.get(activation)
         self.inner_activation = activations.get(inner_activation)
-        self.input = T.tensor3()
+        super(GRU, self).__init__(**kwargs)
 
-        self.W_z = self.init((self.input_dim, self.output_dim))
+    def build(self):
+        input_shape = self.input_shape
+        input_dim = input_shape[2]
+        self.input_dim = input_dim
+        self.input = K.placeholder(input_shape)
+
+        self.W_z = self.init((input_dim, self.output_dim))
         self.U_z = self.inner_init((self.output_dim, self.output_dim))
-        self.b_z = shared_zeros((self.output_dim))
+        self.b_z = K.zeros((self.output_dim,))
 
-        self.W_r = self.init((self.input_dim, self.output_dim))
+        self.W_r = self.init((input_dim, self.output_dim))
         self.U_r = self.inner_init((self.output_dim, self.output_dim))
-        self.b_r = shared_zeros((self.output_dim))
+        self.b_r = K.zeros((self.output_dim,))
 
-        self.W_h = self.init((self.input_dim, self.output_dim))
+        self.W_h = self.init((input_dim, self.output_dim))
         self.U_h = self.inner_init((self.output_dim, self.output_dim))
-        self.b_h = shared_zeros((self.output_dim))
+        self.b_h = K.zeros((self.output_dim,))
 
-        self.params = [
-            self.W_z, self.U_z, self.b_z,
-            self.W_r, self.U_r, self.b_r,
-            self.W_h, self.U_h, self.b_h,
-        ]
+        self.params = [self.W_z, self.U_z, self.b_z,
+                       self.W_r, self.U_r, self.b_r,
+                       self.W_h, self.U_h, self.b_h]
 
-        if weights is not None:
-            self.set_weights(weights)
+        if self.stateful:
+            if not input_shape[0]:
+                raise Exception('If a RNN is stateful, a complete ' +
+                                'input_shape must be provided ' +
+                                '(including batch size).')
+            self.states = [K.zeros(input_shape[0], self.output_dim)]
+        else:
+            # initial states: all-zero tensor of shape (output_dim)
+            self.states = [None]
 
-    def _step(self,
-              xz_t, xr_t, xh_t, mask_tm1,
-              h_tm1,
-              u_z, u_r, u_h):
-        h_mask_tm1 = mask_tm1 * h_tm1
-        z = self.inner_activation(xz_t + T.dot(h_mask_tm1, u_z))
-        r = self.inner_activation(xr_t + T.dot(h_mask_tm1, u_r))
-        hh_t = self.activation(xh_t + T.dot(r * h_mask_tm1, u_h))
-        h_t = z * h_mask_tm1 + (1 - z) * hh_t
-        return h_t
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
 
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
+    def step(self, x, states):
+        assert len(states) == 1
+        x_z = K.dot(x, self.W_z) + self.b_z
+        x_r = K.dot(x, self.W_r) + self.b_r
+        x_h = K.dot(x, self.W_h) + self.b_h
 
-        x_z = T.dot(X, self.W_z) + self.b_z
-        x_r = T.dot(X, self.W_r) + self.b_r
-        x_h = T.dot(X, self.W_h) + self.b_h
-        outputs, updates = theano.scan(
-            self._step,
-            sequences=[x_z, x_r, x_h, padded_mask],
-            outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-            non_sequences=[self.U_z, self.U_r, self.U_h],
-            truncate_gradient=self.truncate_gradient)
+        h_tm1 = states[0]
+        z = self.inner_activation(x_z + K.dot(h_tm1, self.U_z))
+        r = self.inner_activation(x_r + K.dot(h_tm1, self.U_r))
 
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
+        hh = self.inner_activation(x_h + K.dot(r * h_tm1, self.U_h))
+        h = z * h_tm1 + (1 - z) * hh
+        return h, [h]
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
+        config = {"output_dim": self.output_dim,
+                  "init": self.init.__name__,
+                  "inner_init": self.inner_init.__name__,
+                  "activation": self.activation.__name__,
+                  "inner_activation": self.inner_activation.__name__}
+        base_config = super(GRU, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
 
 
 class LSTM(Recurrent):
     '''
         Acts as a spatiotemporal projection,
         turning a sequence of vectors into a single vector.
-
-        Eats inputs with shape:
-        (nb_samples, max_sample_length (samples shorter than this are padded with zeros at the end), input_dim)
-
+        Takes inputs with shape:
+        (nb_samples, max_sample_length, input_dim)
+        (samples shorter than `max_sample_length`
+         are padded with zeros at the end)
         and returns outputs with shape:
         if not return_sequences:
             (nb_samples, output_dim)
         if return_sequences:
             (nb_samples, max_sample_length, output_dim)
-
         For a step-by-step description of the algorithm, see:
         http://deeplearning.net/tutorial/lstm.html
-
         References:
             Long short-term memory (original 97 paper)
                 http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf
@@ -319,407 +272,83 @@ class LSTM(Recurrent):
             Supervised sequence labelling with recurrent neural networks
                 http://www.cs.toronto.edu/~graves/preprint.pdf
     '''
-    def __init__(self, input_dim, output_dim=128,
-                 init='glorot_uniform', inner_init='orthogonal', forget_bias_init='one',
-                 activation='tanh', inner_activation='hard_sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(LSTM, self).__init__()
-        self.input_dim = input_dim
+    def __init__(self, output_dim,
+                 init='glorot_uniform', inner_init='orthogonal',
+                 forget_bias_init='one', activation='tanh',
+                 inner_activation='hard_sigmoid', **kwargs):
         self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.return_sequences = return_sequences
-
         self.init = initializations.get(init)
         self.inner_init = initializations.get(inner_init)
         self.forget_bias_init = initializations.get(forget_bias_init)
         self.activation = activations.get(activation)
         self.inner_activation = activations.get(inner_activation)
-        self.input = T.tensor3()
+        super(LSTM, self).__init__(**kwargs)
 
-        self.W_i = self.init((self.input_dim, self.output_dim))
+    def build(self):
+        input_shape = self.input_shape
+        input_dim = input_shape[2]
+        self.input_dim = input_dim
+        self.input = K.placeholder(input_shape)
+
+        if self.stateful:
+            if not input_shape[0]:
+                raise Exception('If a RNN is stateful, a complete ' +
+                                'input_shape must be provided ' +
+                                '(including batch size).')
+            self.states = [K.zeros(input_shape[0], self.output_dim),
+                           K.zeros(input_shape[0], self.output_dim)]
+        else:
+            # initial states: 2 all-zero tensor of shape (output_dim)
+            self.states = [None, None]
+
+        self.W_i = self.init((input_dim, self.output_dim))
         self.U_i = self.inner_init((self.output_dim, self.output_dim))
-        self.b_i = shared_zeros((self.output_dim))
+        self.b_i = K.zeros((self.output_dim))
 
-        self.W_f = self.init((self.input_dim, self.output_dim))
+        self.W_f = self.init((input_dim, self.output_dim))
         self.U_f = self.inner_init((self.output_dim, self.output_dim))
         self.b_f = self.forget_bias_init((self.output_dim))
 
-        self.W_c = self.init((self.input_dim, self.output_dim))
+        self.W_c = self.init((input_dim, self.output_dim))
         self.U_c = self.inner_init((self.output_dim, self.output_dim))
-        self.b_c = shared_zeros((self.output_dim))
+        self.b_c = K.zeros((self.output_dim))
 
-        self.W_o = self.init((self.input_dim, self.output_dim))
+        self.W_o = self.init((input_dim, self.output_dim))
         self.U_o = self.inner_init((self.output_dim, self.output_dim))
-        self.b_o = shared_zeros((self.output_dim))
+        self.b_o = K.zeros((self.output_dim))
 
-        self.params = [
-            self.W_i, self.U_i, self.b_i,
-            self.W_c, self.U_c, self.b_c,
-            self.W_f, self.U_f, self.b_f,
-            self.W_o, self.U_o, self.b_o,
-        ]
+        self.params = [self.W_i, self.U_i, self.b_i,
+                       self.W_c, self.U_c, self.b_c,
+                       self.W_f, self.U_f, self.b_f,
+                       self.W_o, self.U_o, self.b_o]
 
-        if weights is not None:
-            self.set_weights(weights)
+        if self.initial_weights is not None:
+            self.set_weights(self.initial_weights)
+            del self.initial_weights
 
-    def _step(self,
-              xi_t, xf_t, xo_t, xc_t, mask_tm1,
-              h_tm1, c_tm1,
-              u_i, u_f, u_o, u_c):
-        h_mask_tm1 = mask_tm1 * h_tm1
-        c_mask_tm1 = mask_tm1 * c_tm1
+    def step(self, x, states):
+        assert len(states) == 2
+        h_tm1 = states[0]
+        c_tm1 = states[1]
 
-        i_t = self.inner_activation(xi_t + T.dot(h_mask_tm1, u_i))
-        f_t = self.inner_activation(xf_t + T.dot(h_mask_tm1, u_f))
-        c_t = f_t * c_mask_tm1 + i_t * self.activation(xc_t + T.dot(h_mask_tm1, u_c))
-        o_t = self.inner_activation(xo_t + T.dot(h_mask_tm1, u_o))
-        h_t = o_t * self.activation(c_t)
-        return h_t, c_t
+        x_i = K.dot(x, self.W_i) + self.b_i
+        x_f = K.dot(x, self.W_f) + self.b_f
+        x_c = K.dot(x, self.W_c) + self.b_c
+        x_o = K.dot(x, self.W_o) + self.b_o
 
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
-
-        xi = T.dot(X, self.W_i) + self.b_i
-        xf = T.dot(X, self.W_f) + self.b_f
-        xc = T.dot(X, self.W_c) + self.b_c
-        xo = T.dot(X, self.W_o) + self.b_o
-
-        [outputs, memories], updates = theano.scan(
-            self._step,
-            sequences=[xi, xf, xo, xc, padded_mask],
-            outputs_info=[
-                T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-                T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1)
-            ],
-            non_sequences=[self.U_i, self.U_f, self.U_o, self.U_c],
-            truncate_gradient=self.truncate_gradient)
-
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
+        i = self.inner_activation(x_i + K.dot(h_tm1, self.U_i))
+        f = self.inner_activation(x_f + K.dot(h_tm1, self.U_f))
+        c = f * c_tm1 + i * self.activation(x_c + K.dot(h_tm1, self.U_c))
+        o = self.inner_activation(x_o + K.dot(h_tm1, self.U_o))
+        h = o * self.activation(c)
+        return h, [h, c]
 
     def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "forget_bias_init": self.forget_bias_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
-
-
-class JZS1(Recurrent):
-    '''
-        Evolved recurrent neural network architectures from the evaluation of thousands
-        of models, serving as alternatives to LSTMs and GRUs. See Jozefowicz et al. 2015.
-
-        This corresponds to the `MUT1` architecture described in the paper.
-
-        Takes inputs with shape:
-        (nb_samples, max_sample_length (samples shorter than this are padded with zeros at the end), input_dim)
-
-        and returns outputs with shape:
-        if not return_sequences:
-            (nb_samples, output_dim)
-        if return_sequences:
-            (nb_samples, max_sample_length, output_dim)
-
-        References:
-            An Empirical Exploration of Recurrent Network Architectures
-                http://www.jmlr.org/proceedings/papers/v37/jozefowicz15.pdf
-    '''
-    def __init__(self, input_dim, output_dim=128,
-                 init='glorot_uniform', inner_init='orthogonal',
-                 activation='tanh', inner_activation='sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(JZS1, self).__init__()
-        self.input_dim = input_dim
-        self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.return_sequences = return_sequences
-
-        self.init = initializations.get(init)
-        self.inner_init = initializations.get(inner_init)
-        self.activation = activations.get(activation)
-        self.inner_activation = activations.get(inner_activation)
-        self.input = T.tensor3()
-
-        self.W_z = self.init((self.input_dim, self.output_dim))
-        self.b_z = shared_zeros((self.output_dim))
-
-        self.W_r = self.init((self.input_dim, self.output_dim))
-        self.U_r = self.inner_init((self.output_dim, self.output_dim))
-        self.b_r = shared_zeros((self.output_dim))
-
-        self.U_h = self.inner_init((self.output_dim, self.output_dim))
-        self.b_h = shared_zeros((self.output_dim))
-
-        # P_h used to project X onto different dimension, using sparse random projections
-        if self.input_dim == self.output_dim:
-            self.Pmat = theano.shared(np.identity(self.output_dim, dtype=theano.config.floatX), name=None)
-        else:
-            P = np.random.binomial(1, 0.5, size=(self.input_dim, self.output_dim)).astype(theano.config.floatX) * 2 - 1
-            P = 1 / np.sqrt(self.input_dim) * P
-            self.Pmat = theano.shared(P, name=None)
-
-        self.params = [
-            self.W_z, self.b_z,
-            self.W_r, self.U_r, self.b_r,
-            self.U_h, self.b_h,
-        ]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def _step(self,
-              xz_t, xr_t, xh_t, mask_tm1,
-              h_tm1,
-              u_r, u_h):
-        h_mask_tm1 = mask_tm1 * h_tm1
-        z = self.inner_activation(xz_t)
-        r = self.inner_activation(xr_t + T.dot(h_mask_tm1, u_r))
-        hh_t = self.activation(xh_t + T.dot(r * h_mask_tm1, u_h))
-        h_t = hh_t * z + h_mask_tm1 * (1 - z)
-        return h_t
-
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
-
-        x_z = T.dot(X, self.W_z) + self.b_z
-        x_r = T.dot(X, self.W_r) + self.b_r
-        x_h = T.tanh(T.dot(X, self.Pmat)) + self.b_h
-        outputs, updates = theano.scan(
-            self._step,
-            sequences=[x_z, x_r, x_h, padded_mask],
-            outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-            non_sequences=[self.U_r, self.U_h],
-            truncate_gradient=self.truncate_gradient)
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
-
-
-class JZS2(Recurrent):
-    '''
-        Evolved recurrent neural network architectures from the evaluation of thousands
-        of models, serving as alternatives to LSTMs and GRUs. See Jozefowicz et al. 2015.
-
-        This corresponds to the `MUT2` architecture described in the paper.
-
-        Takes inputs with shape:
-        (nb_samples, max_sample_length (samples shorter than this are padded with zeros at the end), input_dim)
-
-        and returns outputs with shape:
-        if not return_sequences:
-            (nb_samples, output_dim)
-        if return_sequences:
-            (nb_samples, max_sample_length, output_dim)
-
-        References:
-            An Empirical Exploration of Recurrent Network Architectures
-                http://www.jmlr.org/proceedings/papers/v37/jozefowicz15.pdf
-    '''
-    def __init__(self, input_dim, output_dim=128,
-                 init='glorot_uniform', inner_init='orthogonal',
-                 activation='tanh', inner_activation='sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(JZS2, self).__init__()
-        self.input_dim = input_dim
-        self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.return_sequences = return_sequences
-
-        self.init = initializations.get(init)
-        self.inner_init = initializations.get(inner_init)
-        self.activation = activations.get(activation)
-        self.inner_activation = activations.get(inner_activation)
-        self.input = T.tensor3()
-
-        self.W_z = self.init((self.input_dim, self.output_dim))
-        self.U_z = self.inner_init((self.output_dim, self.output_dim))
-        self.b_z = shared_zeros((self.output_dim))
-
-        self.U_r = self.inner_init((self.output_dim, self.output_dim))
-        self.b_r = shared_zeros((self.output_dim))
-
-        self.W_h = self.init((self.input_dim, self.output_dim))
-        self.U_h = self.inner_init((self.output_dim, self.output_dim))
-        self.b_h = shared_zeros((self.output_dim))
-
-        # P_h used to project X onto different dimension, using sparse random projections
-        if self.input_dim == self.output_dim:
-            self.Pmat = theano.shared(np.identity(self.output_dim, dtype=theano.config.floatX), name=None)
-        else:
-            P = np.random.binomial(1, 0.5, size=(self.input_dim, self.output_dim)).astype(theano.config.floatX) * 2 - 1
-            P = 1 / np.sqrt(self.input_dim) * P
-            self.Pmat = theano.shared(P, name=None)
-
-        self.params = [
-            self.W_z, self.U_z, self.b_z,
-            self.U_r, self.b_r,
-            self.W_h, self.U_h, self.b_h,
-        ]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def _step(self,
-              xz_t, xr_t, xh_t, mask_tm1,
-              h_tm1,
-              u_z, u_r, u_h):
-        h_mask_tm1 = mask_tm1 * h_tm1
-        z = self.inner_activation(xz_t + T.dot(h_mask_tm1, u_z))
-        r = self.inner_activation(xr_t + T.dot(h_mask_tm1, u_r))
-        hh_t = self.activation(xh_t + T.dot(r * h_mask_tm1, u_h))
-        h_t = hh_t * z + h_mask_tm1 * (1 - z)
-        return h_t
-
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
-
-        x_z = T.dot(X, self.W_z) + self.b_z
-        x_r = T.dot(X, self.Pmat) + self.b_r
-        x_h = T.dot(X, self.W_h) + self.b_h
-        outputs, updates = theano.scan(
-            self._step,
-            sequences=[x_z, x_r, x_h, padded_mask],
-            outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-            non_sequences=[self.U_z, self.U_r, self.U_h],
-            truncate_gradient=self.truncate_gradient)
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
-
-
-class JZS3(Recurrent):
-    '''
-        Evolved recurrent neural network architectures from the evaluation of thousands
-        of models, serving as alternatives to LSTMs and GRUs. See Jozefowicz et al. 2015.
-
-        This corresponds to the `MUT3` architecture described in the paper.
-
-        Takes inputs with shape:
-        (nb_samples, max_sample_length (samples shorter than this are padded with zeros at the end), input_dim)
-
-        and returns outputs with shape:
-        if not return_sequences:
-            (nb_samples, output_dim)
-        if return_sequences:
-            (nb_samples, max_sample_length, output_dim)
-
-        References:
-            An Empirical Exploration of Recurrent Network Architectures
-                http://www.jmlr.org/proceedings/papers/v37/jozefowicz15.pdf
-    '''
-    def __init__(self, input_dim, output_dim=128,
-                 init='glorot_uniform', inner_init='orthogonal',
-                 activation='tanh', inner_activation='sigmoid',
-                 weights=None, truncate_gradient=-1, return_sequences=False):
-
-        super(JZS3, self).__init__()
-        self.input_dim = input_dim
-        self.output_dim = output_dim
-        self.truncate_gradient = truncate_gradient
-        self.return_sequences = return_sequences
-
-        self.init = initializations.get(init)
-        self.inner_init = initializations.get(inner_init)
-        self.activation = activations.get(activation)
-        self.inner_activation = activations.get(inner_activation)
-        self.input = T.tensor3()
-
-        self.W_z = self.init((self.input_dim, self.output_dim))
-        self.U_z = self.inner_init((self.output_dim, self.output_dim))
-        self.b_z = shared_zeros((self.output_dim))
-
-        self.W_r = self.init((self.input_dim, self.output_dim))
-        self.U_r = self.inner_init((self.output_dim, self.output_dim))
-        self.b_r = shared_zeros((self.output_dim))
-
-        self.W_h = self.init((self.input_dim, self.output_dim))
-        self.U_h = self.inner_init((self.output_dim, self.output_dim))
-        self.b_h = shared_zeros((self.output_dim))
-
-        self.params = [
-            self.W_z, self.U_z, self.b_z,
-            self.W_r, self.U_r, self.b_r,
-            self.W_h, self.U_h, self.b_h,
-        ]
-
-        if weights is not None:
-            self.set_weights(weights)
-
-    def _step(self,
-              xz_t, xr_t, xh_t, mask_tm1,
-              h_tm1,
-              u_z, u_r, u_h):
-        h_mask_tm1 = mask_tm1 * h_tm1
-        z = self.inner_activation(xz_t + T.dot(T.tanh(h_mask_tm1), u_z))
-        r = self.inner_activation(xr_t + T.dot(h_mask_tm1, u_r))
-        hh_t = self.activation(xh_t + T.dot(r * h_mask_tm1, u_h))
-        h_t = hh_t * z + h_mask_tm1 * (1 - z)
-        return h_t
-
-    def get_output(self, train=False):
-        X = self.get_input(train)
-        padded_mask = self.get_padded_shuffled_mask(train, X, pad=1)
-        X = X.dimshuffle((1, 0, 2))
-
-        x_z = T.dot(X, self.W_z) + self.b_z
-        x_r = T.dot(X, self.W_r) + self.b_r
-        x_h = T.dot(X, self.W_h) + self.b_h
-        outputs, updates = theano.scan(
-            self._step,
-            sequences=[x_z, x_r, x_h, padded_mask],
-            outputs_info=T.unbroadcast(alloc_zeros_matrix(X.shape[1], self.output_dim), 1),
-            non_sequences=[self.U_z, self.U_r, self.U_h],
-            truncate_gradient=self.truncate_gradient
-        )
-        if self.return_sequences:
-            return outputs.dimshuffle((1, 0, 2))
-        return outputs[-1]
-
-    def get_config(self):
-        return {"name": self.__class__.__name__,
-                "input_dim": self.input_dim,
-                "output_dim": self.output_dim,
-                "init": self.init.__name__,
-                "inner_init": self.inner_init.__name__,
-                "activation": self.activation.__name__,
-                "inner_activation": self.inner_activation.__name__,
-                "truncate_gradient": self.truncate_gradient,
-                "return_sequences": self.return_sequences}
+        config = {"output_dim": self.output_dim,
+                  "init": self.init.__name__,
+                  "inner_init": self.inner_init.__name__,
+                  "forget_bias_init": self.forget_bias_init.__name__,
+                  "activation": self.activation.__name__,
+                  "inner_activation": self.inner_activation.__name__}
+        base_config = super(LSTM, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))

keras/models.py

@@ -1,12 +1,14 @@
 from __future__ import absolute_import
 from __future__ import print_function
-import theano
-import theano.tensor as T
 import numpy as np
-import warnings, time, copy, pprint
+import warnings
+import time
+import copy
+import pprint
 from six.moves import range
 import six
 
+from . import backend as K
 from . import optimizers
 from . import objectives
 from . import regularizers
@@ -26,11 +28,11 @@ def standardize_y(y):
 
 
 def batch_shuffle(index_array, batch_size):
-    batch_count = int(len(index_array)/batch_size)
+    batch_count = int(len(index_array) / batch_size)
     # to reshape we need to be cleanly divisible by batch size
     # we stash extra items and reappend them after shuffling
-    last_batch = index_array[batch_count*batch_size:]
-    index_array = index_array[:batch_count*batch_size]
+    last_batch = index_array[batch_count * batch_size:]
+    index_array = index_array[:batch_count * batch_size]
     index_array = index_array.reshape((batch_count, batch_size))
     np.random.shuffle(index_array)
     index_array = index_array.flatten()
@@ -38,8 +40,8 @@ def batch_shuffle(index_array, batch_size):
 
 
 def make_batches(size, batch_size):
-    nb_batch = int(np.ceil(size/float(batch_size)))
-    return [(i*batch_size, min(size, (i+1)*batch_size)) for i in range(0, nb_batch)]
+    nb_batch = int(np.ceil(size / float(batch_size)))
+    return [(i * batch_size, min(size, (i + 1) * batch_size)) for i in range(0, nb_batch)]
 
 
 def standardize_X(X):
@@ -52,11 +54,16 @@ def standardize_X(X):
 def slice_X(X, start=None, stop=None):
     if type(X) == list:
         if hasattr(start, '__len__'):
+            # hdf5 dataset only support list object as indices
+            if hasattr(start, 'shape'):
+                start = start.tolist()
             return [x[start] for x in X]
         else:
             return [x[start:stop] for x in X]
     else:
         if hasattr(start, '__len__'):
+            if hasattr(start, 'shape'):
+                start = start.tolist()
             return X[start]
         else:
             return X[start:stop]
@@ -64,65 +71,71 @@ def slice_X(X, start=None, stop=None):
 
 def weighted_objective(fn):
     def weighted(y_true, y_pred, weights, mask=None):
-        # it's important that 0 * Inf == 0, not NaN, so we need to filter
-        # those out first
-        filtered_y_true = y_true[weights.nonzero()[:-1]]
-        filtered_y_pred = y_pred[weights.nonzero()[:-1]]
-        filtered_weights = weights[weights.nonzero()]
-        obj_output = fn(filtered_y_true, filtered_y_pred)
-        weighted = filtered_weights * obj_output
-        if mask is None:
-            # Instead of calling mean() here, we divide by the sum of filtered_weights.
-            return weighted.sum() / filtered_weights.sum()
-        else:
-            filtered_mask = mask[weights.nonzero()[:-1]]
-            return weighted.sum() / (filtered_mask * filtered_weights).sum()
+        '''To be called only with non-zero weights.
+
+        mask: binary
+        '''
+        score_array = fn(y_true, y_pred)
+        if mask is not None:
+            score_array *= mask
+            #  the loss per batch should be proportional
+            #  to the number of unmasked sampled.
+            score_array /= K.mean(mask)
+
+        # reduce score_array to 1D
+        ndim = K.ndim(score_array)
+        for d in range(ndim-1):
+            score_array = K.mean(score_array, axis=-1)
+
+        if weights is not None:
+            score_array *= weights
+        return K.mean(score_array)
     return weighted
 
 
 def standardize_weights(y, sample_weight=None, class_weight=None):
     if sample_weight is not None:
-        return standardize_y(sample_weight)
+        assert len(sample_weight) == len(y)
+        return sample_weight.flatten()
     elif isinstance(class_weight, dict):
-        if len(y.shape) > 3:
-            raise Exception('class_weight not supported for 4+ dimensional targets.')
-        yshape = y.shape
-        y = np.reshape(y, (-1, yshape[-1]))  # for time-distributed data, collapse time and sample
+        if len(y.shape) > 2:
+            raise Exception('class_weight not supported for 3+ dimensional targets.')
         if y.shape[1] > 1:
             y_classes = y.argmax(axis=1)
         elif y.shape[1] == 1:
             y_classes = np.reshape(y, y.shape[0])
         else:
             y_classes = y
-        class_weights = np.asarray([class_weight[cls] for cls in y_classes])
-        return np.reshape(class_weights, yshape[:-1] + (1,))  # uncollapse initial dimensions
+        weights = np.asarray([class_weight[cls] for cls in y_classes])
+        return weights
     else:
-        return np.ones(y.shape[:-1] + (1,))
+        return np.ones((y.shape[0],))
 
 
-def model_from_yaml(yaml_string):
+def model_from_yaml(yaml_string, custom_objects={}):
     '''
         Returns a model generated from a local yaml file,
-        which is either created by hand or from to_yaml method of Sequential or Graph
+        which is either created by hand or from to_yaml method
+        of Sequential or Graph
     '''
     import yaml
     config = yaml.load(yaml_string)
-    return model_from_config(config)
+    return model_from_config(config, custom_objects=custom_objects)
 
 
-def model_from_json(json_string):
+def model_from_json(json_string, custom_objects={}):
     import json
     config = json.loads(json_string)
-    return model_from_config(config)
+    return model_from_config(config, custom_objects=custom_objects)
 
 
-def model_from_config(config):
+def model_from_config(config, custom_objects={}):
     model_name = config.get('name')
     if model_name not in {'Graph', 'Sequential'}:
         raise Exception('Unrecognized model:', model_name)
 
     # Create a container then set class to appropriate model
-    model = container_from_config(config)
+    model = container_from_config(config, custom_objects=custom_objects)
     if model_name == 'Graph':
         model.__class__ = Graph
     elif model_name == 'Sequential':
@@ -139,10 +152,11 @@ def model_from_config(config):
         optimizer = optimizers.get(optimizer_name, optimizer_params)
 
         if model_name == 'Sequential':
-            model.compile(loss=loss, optimizer=optimizer, class_mode=class_mode, theano_mode=theano_mode)
+            model.compile(loss=loss, optimizer=optimizer,
+                          class_mode=class_mode, theano_mode=theano_mode)
         elif model_name == 'Graph':
-            model.compile(loss=loss, optimizer=optimizer, theano_mode=theano_mode)
-
+            model.compile(loss=loss, optimizer=optimizer,
+                          theano_mode=theano_mode)
     return model
 
 
@@ -154,10 +168,12 @@ def get_function_name(o):
 
 
 class Model(object):
-    def _fit(self, f, ins, out_labels=[], batch_size=128, nb_epoch=100, verbose=1, callbacks=[],
+    def _fit(self, f, ins, out_labels=[], batch_size=128,
+             nb_epoch=100, verbose=1, callbacks=[],
              val_f=None, val_ins=None, shuffle=True, metrics=[]):
         '''
-            Abstract fit function for f(*ins). Assume that f returns a list, labelled by out_labels.
+            Abstract fit function for f(ins).
+            Assume that f returns a list, labelled by out_labels.
         '''
         do_validation = False
         if val_f and val_ins:
@@ -199,16 +215,15 @@ class Model(object):
                 batch_ids = index_array[batch_start:batch_end]
                 try:
                     ins_batch = slice_X(ins, batch_ids)
-                except TypeError as err:
-                    print('TypeError while preparing batch. \
+                except TypeError:
+                    raise Exception('TypeError while preparing batch. \
                         If using HDF5 input data, pass shuffle="batch".\n')
-                    raise
 
                 batch_logs = {}
                 batch_logs['batch'] = batch_index
                 batch_logs['size'] = len(batch_ids)
                 callbacks.on_batch_begin(batch_index, batch_logs)
-                outs = f(*ins_batch)
+                outs = f(ins_batch)
                 if type(outs) != list:
                     outs = [outs]
                 for l, o in zip(out_labels, outs):
@@ -221,7 +236,9 @@ class Model(object):
                     # validation
                     if do_validation:
                         # replace with self._evaluate
-                        val_outs = self._test_loop(val_f, val_ins, batch_size=batch_size, verbose=0)
+                        val_outs = self._test_loop(val_f, val_ins,
+                                                   batch_size=batch_size,
+                                                   verbose=0)
                         if type(val_outs) != list:
                             val_outs = [val_outs]
                         # same labels assumed
@@ -249,7 +266,7 @@ class Model(object):
             batch_ids = index_array[batch_start:batch_end]
             ins_batch = slice_X(ins, batch_ids)
 
-            batch_outs = f(*ins_batch)
+            batch_outs = f(ins_batch)
             if type(batch_outs) != list:
                 batch_outs = [batch_outs]
             if batch_index == 0:
@@ -277,7 +294,7 @@ class Model(object):
             batch_ids = index_array[batch_start:batch_end]
             ins_batch = slice_X(ins, batch_ids)
 
-            batch_outs = f(*ins_batch)
+            batch_outs = f(ins_batch)
             if type(batch_outs) == list:
                 if batch_index == 0:
                     for batch_out in enumerate(batch_outs):
@@ -313,17 +330,17 @@ class Model(object):
             pp.pprint(config)
         return config
 
-    def to_yaml(self):
+    def to_yaml(self, **kwargs):
         # dump model configuration to yaml string
         import yaml
         config = self.get_config()
-        return yaml.dump(config)
+        return yaml.dump(config, **kwargs)
 
-    def to_json(self):
+    def to_json(self, **kwargs):
         # dump model configuration to json string
         import json
         config = self.get_config()
-        return json.dumps(config)
+        return json.dumps(config, **kwargs)
 
 
 class Sequential(Model, containers.Sequential):
@@ -341,7 +358,8 @@ class Sequential(Model, containers.Sequential):
             - set_weights
     '''
 
-    def compile(self, optimizer, loss, class_mode="categorical", theano_mode=None):
+    def compile(self, optimizer, loss,
+                class_mode="categorical", theano_mode=None):
         self.optimizer = optimizers.get(optimizer)
 
         self.loss = objectives.get(loss)
@@ -355,9 +373,9 @@ class Sequential(Model, containers.Sequential):
         self.y_test = self.get_output(train=False)
 
         # target of model
-        self.y = T.zeros_like(self.y_train)
-
-        self.weights = T.ones_like(self.y_train)
+        self.y = K.placeholder(ndim=K.ndim(self.y_train))
+        # weights: one scalar per sample
+        self.weights = K.placeholder(ndim=1)
 
         if hasattr(self.layers[-1], "get_output_mask"):
             mask = self.layers[-1].get_output_mask()
@@ -366,17 +384,15 @@ class Sequential(Model, containers.Sequential):
         train_loss = weighted_loss(self.y, self.y_train, self.weights, mask)
         test_loss = weighted_loss(self.y, self.y_test, self.weights, mask)
 
-        train_loss.name = 'train_loss'
-        test_loss.name = 'test_loss'
-        self.y.name = 'y'
-
         if class_mode == "categorical":
-            train_accuracy = T.mean(T.eq(T.argmax(self.y, axis=-1), T.argmax(self.y_train, axis=-1)))
-            test_accuracy = T.mean(T.eq(T.argmax(self.y, axis=-1), T.argmax(self.y_test, axis=-1)))
+            train_accuracy = K.mean(K.equal(K.argmax(self.y, axis=-1),
+                                            K.argmax(self.y_train, axis=-1)))
+            test_accuracy = K.mean(K.equal(K.argmax(self.y, axis=-1),
+                                           K.argmax(self.y_test, axis=-1)))
 
         elif class_mode == "binary":
-            train_accuracy = T.mean(T.eq(self.y, T.round(self.y_train)))
-            test_accuracy = T.mean(T.eq(self.y, T.round(self.y_test)))
+            train_accuracy = K.mean(K.equal(self.y, K.round(self.y_train)))
+            test_accuracy = K.mean(K.equal(self.y, K.round(self.y_test)))
         else:
             raise Exception("Invalid class mode:" + str(class_mode))
         self.class_mode = class_mode
@@ -384,39 +400,38 @@ class Sequential(Model, containers.Sequential):
 
         for r in self.regularizers:
             train_loss = r(train_loss)
-        updates = self.optimizer.get_updates(self.params, self.constraints, train_loss)
+        updates = self.optimizer.get_updates(self.params,
+                                             self.constraints,
+                                             train_loss)
         updates += self.updates
 
         if type(self.X_train) == list:
             train_ins = self.X_train + [self.y, self.weights]
             test_ins = self.X_test + [self.y, self.weights]
+            assert type(self.X_test) == list
             predict_ins = self.X_test
         else:
             train_ins = [self.X_train, self.y, self.weights]
             test_ins = [self.X_test, self.y, self.weights]
             predict_ins = [self.X_test]
 
-        self._train = theano.function(train_ins, train_loss, updates=updates,
-                                      allow_input_downcast=True, mode=theano_mode)
-        self._train_with_acc = theano.function(train_ins, [train_loss, train_accuracy], updates=updates,
-                                               allow_input_downcast=True, mode=theano_mode)
-        self._predict = theano.function(predict_ins, self.y_test,
-                                        allow_input_downcast=True, mode=theano_mode)
-        self._test = theano.function(test_ins, test_loss,
-                                     allow_input_downcast=True, mode=theano_mode)
-        self._test_with_acc = theano.function(test_ins, [test_loss, test_accuracy],
-                                              allow_input_downcast=True, mode=theano_mode)
+        self._train = K.function(train_ins, [train_loss], updates=updates)
+        self._train_with_acc = K.function(train_ins, [train_loss, train_accuracy], updates=updates)
+        self._predict = K.function(predict_ins, [self.y_test])
+        self._test = K.function(test_ins, [test_loss])
+        self._test_with_acc = K.function(test_ins, [test_loss, test_accuracy])
 
-    def train_on_batch(self, X, y, accuracy=False, class_weight=None, sample_weight=None):
+    def train_on_batch(self, X, y, accuracy=False,
+                       class_weight=None, sample_weight=None):
         X = standardize_X(X)
         y = standardize_y(y)
-        sample_weight = standardize_weights(y, class_weight=class_weight, sample_weight=sample_weight)
-
+        sample_weight = standardize_weights(y, class_weight=class_weight,
+                                            sample_weight=sample_weight)
         ins = X + [y, sample_weight]
         if accuracy:
-            return self._train_with_acc(*ins)
+            return self._train_with_acc(ins)
         else:
-            return self._train(*ins)
+            return self._train(ins)
 
     def test_on_batch(self, X, y, accuracy=False, sample_weight=None):
         X = standardize_X(X)
@@ -425,17 +440,17 @@ class Sequential(Model, containers.Sequential):
 
         ins = X + [y, sample_weight]
         if accuracy:
-            return self._test_with_acc(*ins)
+            return self._test_with_acc(ins)
         else:
-            return self._test(*ins)
+            return self._test(ins)
 
     def predict_on_batch(self, X):
         ins = standardize_X(X)
-        return self._predict(*ins)
+        return self._predict(ins)
 
     def fit(self, X, y, batch_size=128, nb_epoch=100, verbose=1, callbacks=[],
-            validation_split=0., validation_data=None, shuffle=True, show_accuracy=False,
-            class_weight=None, sample_weight=None):
+            validation_split=0., validation_data=None, shuffle=True,
+            show_accuracy=False, class_weight=None, sample_weight=None):
 
         X = standardize_X(X)
         y = standardize_y(y)
@@ -452,12 +467,13 @@ class Sequential(Model, containers.Sequential):
                 X_val, y_val = validation_data
                 X_val = standardize_X(X_val)
                 y_val = standardize_y(y_val)
-                sample_weight_val = np.ones(y_val.shape[:-1] + (1,))
+                sample_weight_val = standardize_weights(y_val)
             elif len(validation_data) == 3:
                 X_val, y_val, sample_weight_val = validation_data
                 X_val = standardize_X(X_val)
                 y_val = standardize_y(y_val)
-                sample_weight_val = standardize_weights(y_val, sample_weight=sample_weight_val)
+                sample_weight_val = standardize_weights(y_val,
+                                                        sample_weight=sample_weight_val)
             else:
                 raise Exception("Invalid format for validation data; provide a tuple (X_val, y_val) or (X_val, y_val, sample_weight). \
                     X_val may be a numpy array or a list of numpy arrays depending on your model input.")
@@ -471,7 +487,7 @@ class Sequential(Model, containers.Sequential):
                 sample_weight, sample_weight_val = (slice_X(sample_weight, 0, split_at), slice_X(sample_weight, split_at))
                 sample_weight_val = standardize_weights(y_val, sample_weight=sample_weight_val)
             else:
-                sample_weight_val = np.ones(y_val.shape[:-1] + (1,))
+                sample_weight_val = standardize_weights(y_val)
             val_ins = X_val + [y_val, sample_weight_val]
 
         if show_accuracy:
@@ -484,7 +500,8 @@ class Sequential(Model, containers.Sequential):
         sample_weight = standardize_weights(y, class_weight=class_weight, sample_weight=sample_weight)
         ins = X + [y, sample_weight]
         metrics = ['loss', 'acc', 'val_loss', 'val_acc']
-        return self._fit(f, ins, out_labels=out_labels, batch_size=batch_size, nb_epoch=nb_epoch,
+        return self._fit(f, ins, out_labels=out_labels,
+                         batch_size=batch_size, nb_epoch=nb_epoch,
                          verbose=verbose, callbacks=callbacks,
                          val_f=val_f, val_ins=val_ins,
                          shuffle=shuffle, metrics=metrics)
@@ -506,7 +523,8 @@ class Sequential(Model, containers.Sequential):
         else:
             return (proba > 0.5).astype('int32')
 
-    def evaluate(self, X, y, batch_size=128, show_accuracy=False, verbose=1, sample_weight=None):
+    def evaluate(self, X, y, batch_size=128, show_accuracy=False,
+                 verbose=1, sample_weight=None):
         X = standardize_X(X)
         y = standardize_y(y)
         sample_weight = standardize_weights(y, sample_weight=sample_weight)
@@ -581,7 +599,7 @@ class Graph(Model, containers.Graph):
             output = self.outputs[output_name]
             y_train = output.get_output(True)
             y_test = output.get_output(False)
-            y = T.zeros_like(y_test)
+            y = K.placeholder(ndim=K.ndim(y_train))
             ys.append(y)
             ys_train.append(y_train)
             ys_test.append(y_test)
@@ -591,15 +609,12 @@ class Graph(Model, containers.Graph):
             else:
                 mask = None
 
-            weight = T.ones_like(y_test)
+            weight = K.placeholder(ndim=1)
             weights.append(weight)
             weighted_loss = weighted_objective(objectives.get(loss_fn))
             train_loss += weighted_loss(y, y_train, weight, mask)
             test_loss += weighted_loss(y, y_test, weight, mask)
 
-        train_loss.name = 'train_loss'
-        test_loss.name = 'test_loss'
-
         ins = [self.inputs[name].input for name in self.input_order]
         train_ins = ins + ys + weights
         test_ins = ins + ys + weights
@@ -612,12 +627,9 @@ class Graph(Model, containers.Graph):
         self.theano_mode = theano_mode
         self.loss = loss
 
-        self._train = theano.function(train_ins, train_loss, updates=updates,
-                                      allow_input_downcast=True, mode=theano_mode)
-        self._test = theano.function(test_ins, test_loss,
-                                     allow_input_downcast=True, mode=theano_mode)
-        self._predict = theano.function(inputs=ins, outputs=ys_test,
-                                        allow_input_downcast=True, mode=theano_mode)
+        self._train = K.function(train_ins, [train_loss], updates=updates)
+        self._test = K.function(test_ins, [test_loss])
+        self._predict = K.function(inputs=ins, outputs=ys_test)
 
     def train_on_batch(self, data, class_weight={}, sample_weight={}):
         # data is a dictionary mapping output and input names to arrays
@@ -625,26 +637,28 @@ class Graph(Model, containers.Graph):
                                              sample_weight=sample_weight.get(name),
                                              class_weight=class_weight.get(name)) for name in self.output_order]
         ins = [data[name] for name in self.input_order] + [standardize_y(data[name]) for name in self.output_order] + sample_weight
-        return self._train(*ins)
+        return self._train(ins)
 
     def test_on_batch(self, data, sample_weight={}):
         # data is a dictionary mapping input names to arrays
         sample_weight = [standardize_weights(data[name],
                                              sample_weight=sample_weight.get(name)) for name in self.output_order]
         ins = [data[name] for name in self.input_order] + [standardize_y(data[name]) for name in self.output_order] + sample_weight
-        return self._test(*ins)
+        return self._test(ins)
 
     def predict_on_batch(self, data):
         # data is a dictionary mapping input names to arrays
         ins = [data[name] for name in self.input_order]
-        return self._predict(*ins)
+        return self._predict(ins)
 
     def fit(self, data, batch_size=128, nb_epoch=100, verbose=1, callbacks=[],
-            validation_split=0., validation_data=None, shuffle=True, class_weight={}, sample_weight={}):
+            validation_split=0., validation_data=None, shuffle=True,
+            class_weight={}, sample_weight={}):
         X = [data[name] for name in self.input_order]
         y = [standardize_y(data[name]) for name in self.output_order]
-        sample_weight_list = [standardize_weights(data[name],
-                                                  sample_weight=sample_weight.get(name)) for name in self.output_order]
+
+        sample_weight_list = [standardize_weights(y[i],
+                                                  sample_weight=sample_weight.get(self.output_order[i])) for i in range(len(self.output_order))]
         class_weight_list = [class_weight.get(name) for name in self.output_order]
 
         val_f = None
@@ -653,7 +667,8 @@ class Graph(Model, containers.Graph):
             val_f = self._test
         if validation_data:
             # can't use sample weights with validation data at this point
-            sample_weight = [standardize_weights(validation_data[name]) for name in self.output_order]
+            y_val = [standardize_y(data[name]) for name in self.output_order]
+            sample_weight = [standardize_weights(y_val[i]) for i in range(len(y_val))]
             val_ins = [validation_data[name] for name in self.input_order] + [standardize_y(validation_data[name]) for name in self.output_order] + sample_weight
 
         elif 0 < validation_split < 1:
@@ -671,7 +686,6 @@ class Graph(Model, containers.Graph):
                                                   sample_weight=sample_weight_list[i],
                                                   class_weight=class_weight_list[i]) for i in range(len(self.output_order))]
         ins = X + y + sample_weight_list
-
         history = self._fit(f, ins, out_labels=out_labels, batch_size=batch_size, nb_epoch=nb_epoch,
                             verbose=verbose, callbacks=callbacks,
                             val_f=val_f, val_ins=val_ins,

keras/objectives.py

@@ -1,60 +1,55 @@
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
 import numpy as np
-from six.moves import range
-
-if theano.config.floatX == 'float64':
-    epsilon = 1.0e-9
-else:
-    epsilon = 1.0e-7
+from . import backend as K
 
 
 def mean_squared_error(y_true, y_pred):
-    return T.sqr(y_pred - y_true).mean(axis=-1)
+    return K.mean(K.square(y_pred - y_true), axis=-1)
+
+
+def root_mean_squared_error(y_true, y_pred):
+    return K.sqrt(K.mean(K.square(y_pred - y_true), axis=-1))
 
 
 def mean_absolute_error(y_true, y_pred):
-    return T.abs_(y_pred - y_true).mean(axis=-1)
+    return K.mean(K.abs(y_pred - y_true), axis=-1)
 
 
 def mean_absolute_percentage_error(y_true, y_pred):
-    return T.abs_((y_true - y_pred) / T.clip(T.abs_(y_true), epsilon, np.inf)).mean(axis=-1) * 100.
+    diff = K.abs((y_true - y_pred) / K.clip(K.abs(y_true), K._EPSILON, np.inf))
+    return 100. * K.mean(diff, axis=-1)
 
 
 def mean_squared_logarithmic_error(y_true, y_pred):
-    return T.sqr(T.log(T.clip(y_pred, epsilon, np.inf) + 1.) - T.log(T.clip(y_true, epsilon, np.inf) + 1.)).mean(axis=-1)
+    first_log = K.log(K.clip(y_pred, K._EPSILON, np.inf) + 1.)
+    second_log = K.log(K.clip(y_true, K._EPSILON, np.inf) + 1.)
+    return K.mean(K.square(first_log - second_log), axis=-1)
 
 
 def squared_hinge(y_true, y_pred):
-    return T.sqr(T.maximum(1. - y_true * y_pred, 0.)).mean(axis=-1)
+    return K.mean(K.square(K.maximum(1. - y_true * y_pred, 0.)), axis=-1)
 
 
 def hinge(y_true, y_pred):
-    return T.maximum(1. - y_true * y_pred, 0.).mean(axis=-1)
+    return K.mean(K.maximum(1. - y_true * y_pred, 0.), axis=-1)
 
 
 def categorical_crossentropy(y_true, y_pred):
     '''Expects a binary class matrix instead of a vector of scalar classes
     '''
-    y_pred = T.clip(y_pred, epsilon, 1.0 - epsilon)
-    # scale preds so that the class probas of each sample sum to 1
-    y_pred /= y_pred.sum(axis=-1, keepdims=True)
-    cce = T.nnet.categorical_crossentropy(y_pred, y_true)
-    return cce
+    return K.mean(K.categorical_crossentropy(y_pred, y_true), axis=-1)
 
 
 def binary_crossentropy(y_true, y_pred):
-    y_pred = T.clip(y_pred, epsilon, 1.0 - epsilon)
-    bce = T.nnet.binary_crossentropy(y_pred, y_true).mean(axis=-1)
-    return bce
+    return K.mean(K.binary_crossentropy(y_pred, y_true), axis=-1)
 
 
 def poisson_loss(y_true, y_pred):
-    return T.mean(y_pred - y_true * T.log(y_pred + epsilon), axis=-1)
+    return K.mean(y_pred - y_true * K.log(y_pred + K._EPSILON), axis=-1)
 
 # aliases
 mse = MSE = mean_squared_error
+rmse = RMSE = root_mean_squared_error
 mae = MAE = mean_absolute_error
 mape = MAPE = mean_absolute_percentage_error
 msle = MSLE = mean_squared_logarithmic_error

keras/optimizers.py

@@ -1,20 +1,18 @@
 from __future__ import absolute_import
-import theano
-import theano.tensor as T
-
-from .utils.theano_utils import shared_zeros, shared_scalar, floatX
+from . import backend as K
+import numpy as np
 from .utils.generic_utils import get_from_module
 from six.moves import zip
 
 
 def clip_norm(g, c, n):
     if c > 0:
-        g = T.switch(T.ge(n, c), g * c / n, g)
+        g = K.switch(n >= c, g * c / n, g)
     return g
 
 
 def kl_divergence(p, p_hat):
-    return p_hat - p + p * T.log(p / p_hat)
+    return p_hat - p + p * K.log(p / p_hat)
 
 
 class Optimizer(object):
@@ -23,24 +21,23 @@ class Optimizer(object):
         self.updates = []
 
     def get_state(self):
-        return [u[0].get_value() for u in self.updates]
+        return [K.get_value(u[0]) for u in self.updates]
 
     def set_state(self, value_list):
         assert len(self.updates) == len(value_list)
         for u, v in zip(self.updates, value_list):
-            u[0].set_value(floatX(v))
+            K.set_value(u[0], v)
 
     def get_updates(self, params, constraints, loss):
         raise NotImplementedError
 
     def get_gradients(self, loss, params):
-
-        grads = T.grad(loss, params)
-
+        grads = K.gradients(loss, params)
         if hasattr(self, 'clipnorm') and self.clipnorm > 0:
-            norm = T.sqrt(sum([T.sum(g ** 2) for g in grads]))
+            norm = K.sqrt(sum([K.sum(K.square(g)) for g in grads]))
             grads = [clip_norm(g, self.clipnorm, norm) for g in grads]
-
+        if hasattr(self, 'clipvalue') and self.clipvalue > 0:
+            grads = [K.clip(g, -self.clipvalue, self.clipvalue) for g in grads]
         return grads
 
     def get_config(self):
@@ -49,12 +46,14 @@ class Optimizer(object):
 
 class SGD(Optimizer):
 
-    def __init__(self, lr=0.01, momentum=0., decay=0., nesterov=False, *args, **kwargs):
+    def __init__(self, lr=0.01, momentum=0., decay=0., nesterov=False,
+                 *args, **kwargs):
         super(SGD, self).__init__(**kwargs)
         self.__dict__.update(locals())
-        self.iterations = shared_scalar(0)
-        self.lr = shared_scalar(lr)
-        self.momentum = shared_scalar(momentum)
+        self.iterations = K.variable(0.)
+        self.lr = K.variable(lr)
+        self.momentum = K.variable(momentum)
+        self.decay = K.variable(decay)
 
     def get_updates(self, params, constraints, loss):
         grads = self.get_gradients(loss, params)
@@ -62,7 +61,7 @@ class SGD(Optimizer):
         self.updates = [(self.iterations, self.iterations + 1.)]
 
         for p, g, c in zip(params, grads, constraints):
-            m = shared_zeros(p.get_value().shape)  # momentum
+            m = K.variable(np.zeros(K.get_value(p).shape))  # momentum
             v = self.momentum * m - lr * g  # velocity
             self.updates.append((m, v))
 
@@ -76,9 +75,9 @@ class SGD(Optimizer):
 
     def get_config(self):
         return {"name": self.__class__.__name__,
-                "lr": float(self.lr.get_value()),
-                "momentum": float(self.momentum.get_value()),
-                "decay": float(self.decay.get_value()),
+                "lr": float(K.get_value(self.lr)),
+                "momentum": float(K.get_value(self.momentum)),
+                "decay": float(K.get_value(self.decay)),
                 "nesterov": self.nesterov}
 
 
@@ -86,26 +85,27 @@ class RMSprop(Optimizer):
     def __init__(self, lr=0.001, rho=0.9, epsilon=1e-6, *args, **kwargs):
         super(RMSprop, self).__init__(**kwargs)
         self.__dict__.update(locals())
-        self.lr = shared_scalar(lr)
-        self.rho = shared_scalar(rho)
+        self.lr = K.variable(lr)
+        self.rho = K.variable(rho)
 
     def get_updates(self, params, constraints, loss):
         grads = self.get_gradients(loss, params)
-        accumulators = [shared_zeros(p.get_value().shape) for p in params]
+        accumulators = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
         self.updates = []
 
         for p, g, a, c in zip(params, grads, accumulators, constraints):
-            new_a = self.rho * a + (1 - self.rho) * g ** 2  # update accumulator
+            # update accumulator
+            new_a = self.rho * a + (1 - self.rho) * K.square(g)
             self.updates.append((a, new_a))
 
-            new_p = p - self.lr * g / T.sqrt(new_a + self.epsilon)
+            new_p = p - self.lr * g / K.sqrt(new_a + self.epsilon)
             self.updates.append((p, c(new_p)))  # apply constraints
         return self.updates
 
     def get_config(self):
         return {"name": self.__class__.__name__,
-                "lr": float(self.lr.get_value()),
-                "rho": float(self.rho.get_value()),
+                "lr": float(K.get_value(self.lr)),
+                "rho": float(K.get_value(self.rho)),
                 "epsilon": self.epsilon}
 
 
@@ -113,23 +113,23 @@ class Adagrad(Optimizer):
     def __init__(self, lr=0.01, epsilon=1e-6, *args, **kwargs):
         super(Adagrad, self).__init__(**kwargs)
         self.__dict__.update(locals())
-        self.lr = shared_scalar(lr)
+        self.lr = K.variable(lr)
 
     def get_updates(self, params, constraints, loss):
         grads = self.get_gradients(loss, params)
-        accumulators = [shared_zeros(p.get_value().shape) for p in params]
+        accumulators = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
         self.updates = []
 
         for p, g, a, c in zip(params, grads, accumulators, constraints):
-            new_a = a + g ** 2  # update accumulator
+            new_a = a + K.square(g)  # update accumulator
             self.updates.append((a, new_a))
-            new_p = p - self.lr * g / T.sqrt(new_a + self.epsilon)
+            new_p = p - self.lr * g / K.sqrt(new_a + self.epsilon)
             self.updates.append((p, c(new_p)))  # apply constraints
         return self.updates
 
     def get_config(self):
         return {"name": self.__class__.__name__,
-                "lr": float(self.lr.get_value()),
+                "lr": float(K.get_value(self.lr)),
                 "epsilon": self.epsilon}
 
 
@@ -140,35 +140,35 @@ class Adadelta(Optimizer):
     def __init__(self, lr=1.0, rho=0.95, epsilon=1e-6, *args, **kwargs):
         super(Adadelta, self).__init__(**kwargs)
         self.__dict__.update(locals())
-        self.lr = shared_scalar(lr)
+        self.lr = K.variable(lr)
 
     def get_updates(self, params, constraints, loss):
         grads = self.get_gradients(loss, params)
-        accumulators = [shared_zeros(p.get_value().shape) for p in params]
-        delta_accumulators = [shared_zeros(p.get_value().shape) for p in params]
+        accumulators = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
+        delta_accumulators = [K.variable(np.zeros(K.get_value(p).shape)) for p in params]
         self.updates = []
 
         for p, g, a, d_a, c in zip(params, grads, accumulators,
                                    delta_accumulators, constraints):
-            new_a = self.rho * a + (1 - self.rho) * g ** 2  # update accumulator
+            # update accumulator
+            new_a = self.rho * a + (1 - self.rho) * K.square(g)
             self.updates.append((a, new_a))
 
             # use the new accumulator and the *old* delta_accumulator
-            update = g * T.sqrt(d_a + self.epsilon) / T.sqrt(new_a +
-                                                             self.epsilon)
+            update = g * K.sqrt(d_a + self.epsilon) / K.sqrt(new_a + self.epsilon)
 
             new_p = p - self.lr * update
             self.updates.append((p, c(new_p)))  # apply constraints
 
             # update delta_accumulator
-            new_d_a = self.rho * d_a + (1 - self.rho) * update ** 2
+            new_d_a = self.rho * d_a + (1 - self.rho) * K.square(update)
             self.updates.append((d_a, new_d_a))
         return self.updates
 
     def get_config(self):
         return {"name": self.__class__.__name__,
-                "lr": float(self.lr.get_value()),
-                "rho": self.rho,
+                "lr": float(K.get_value(self.lr)),
+                "rho": float(K.get_value(self.rho)),
                 "epsilon": self.epsilon}
 
 
@@ -178,26 +178,31 @@ class Adam(Optimizer):
 
         Default parameters follow those provided in the original paper.
     '''
-    def __init__(self, lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8, *args, **kwargs):
+    def __init__(self, lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8,
+                 *args, **kwargs):
         super(Adam, self).__init__(**kwargs)
         self.__dict__.update(locals())
-        self.iterations = shared_scalar(0)
-        self.lr = shared_scalar(lr)
+        self.iterations = K.variable(0)
+        self.lr = K.variable(lr)
+        self.beta_1 = K.variable(beta_1)
+        self.beta_2 = K.variable(beta_2)
 
     def get_updates(self, params, constraints, loss):
         grads = self.get_gradients(loss, params)
         self.updates = [(self.iterations, self.iterations+1.)]
 
         t = self.iterations + 1
-        lr_t = self.lr * T.sqrt(1-self.beta_2**t)/(1-self.beta_1**t)
+        lr_t = self.lr * K.sqrt(1 - K.pow(self.beta_2, t)) / (1 - K.pow(self.beta_1, t))
 
         for p, g, c in zip(params, grads, constraints):
-            m = theano.shared(p.get_value() * 0.)  # zero init of moment
-            v = theano.shared(p.get_value() * 0.)  # zero init of velocity
+            # zero init of moment
+            m = K.variable(np.zeros(K.get_value(p).shape))
+            # zero init of velocity
+            v = K.variable(np.zeros(K.get_value(p).shape))
 
             m_t = (self.beta_1 * m) + (1 - self.beta_1) * g
-            v_t = (self.beta_2 * v) + (1 - self.beta_2) * (g**2)
-            p_t = p - lr_t * m_t / (T.sqrt(v_t) + self.epsilon)
+            v_t = (self.beta_2 * v) + (1 - self.beta_2) * K.square(g)
+            p_t = p - lr_t * m_t / (K.sqrt(v_t) + self.epsilon)
 
             self.updates.append((m, m_t))
             self.updates.append((v, v_t))
@@ -206,9 +211,9 @@ class Adam(Optimizer):
 
     def get_config(self):
         return {"name": self.__class__.__name__,
-                "lr": float(self.lr.get_value()),
-                "beta_1": self.beta_1,
-                "beta_2": self.beta_2,
+                "lr": float(K.get_value(self.lr)),
+                "beta_1": float(K.get_value(self.beta_1)),
+                "beta_2": float(K.get_value(self.beta_2)),
                 "epsilon": self.epsilon}
 
 # aliases
@@ -220,5 +225,5 @@ adam = Adam
 
 
 def get(identifier, kwargs=None):
-    return get_from_module(identifier, globals(), 'optimizer', instantiate=True,
-                           kwargs=kwargs)
+    return get_from_module(identifier, globals(), 'optimizer',
+                           instantiate=True, kwargs=kwargs)

keras/preprocessing/image.py

@@ -1,6 +1,7 @@
 from __future__ import absolute_import
 
 import numpy as np
+import re
 from scipy import ndimage
 from scipy import linalg
 
@@ -103,7 +104,7 @@ def img_to_array(img):
 
 def load_img(path, grayscale=False):
     from PIL import Image
-    img = Image.open(open(path))
+    img = Image.open(path)
     if grayscale:
         img = img.convert('L')
     else: # Assure 3 channel even when loaded image is grayscale
@@ -213,17 +214,14 @@ class ImageDataGenerator(object):
         # channel shifting
         return x
 
-
-    def fit(self, X, 
-            augment=False, # fit on randomly augmented samples
-            rounds=1, # if augment, how many augmentation passes over the data do we use
-            seed=None
-        ):
+    def fit(self, X,
+            augment=False,  # fit on randomly augmented samples
+            rounds=1,  # if augment, how many augmentation passes over the data do we use
+            seed=None):
         '''
             Required for featurewise_center, featurewise_std_normalization and zca_whitening.
         '''
         X = np.copy(X)
-        
         if augment:
             aX = np.zeros(tuple([rounds*X.shape[0]]+list(X.shape)[1:]))
             for r in range(rounds):
@@ -246,5 +244,3 @@ class ImageDataGenerator(object):
             sigma = np.dot(flatX.T, flatX) / flatX.shape[1]
             U, S, V = linalg.svd(sigma)
             self.principal_components = np.dot(np.dot(U, np.diag(1. / np.sqrt(S + fudge))), U.T)
-
-

keras/preprocessing/sequence.py

@@ -7,7 +7,7 @@ from six.moves import range
 def pad_sequences(sequences, maxlen=None, dtype='int32', padding='pre', truncating='pre', value=0.):
     """
         Pad each sequence to the same length: 
-        the length of the longuest sequence.
+        the length of the longest sequence.
 
         If maxlen is provided, any sequence longer
         than maxlen is truncated to maxlen. Truncation happens off either the beginning (default) or
@@ -24,6 +24,8 @@ def pad_sequences(sequences, maxlen=None, dtype='int32', padding='pre', truncati
 
     x = (np.ones((nb_samples, maxlen)) * value).astype(dtype)
     for idx, s in enumerate(sequences):
+        if len(s) == 0:
+            continue # empty list was found
         if truncating == 'pre':
             trunc = s[-maxlen:]
         elif truncating == 'post':

keras/regularizers.py

@@ -1,5 +1,5 @@
 from __future__ import absolute_import
-import theano.tensor as T
+from . import backend as K
 
 
 class Regularizer(object):
@@ -25,8 +25,8 @@ class WeightRegularizer(Regularizer):
         self.p = p
 
     def __call__(self, loss):
-        loss += T.sum(abs(self.p)) * self.l1
-        loss += T.sum(self.p ** 2) * self.l2
+        loss += K.sum(K.abs(self.p)) * self.l1
+        loss += K.sum(K.square(self.p)) * self.l2
         return loss
 
     def get_config(self):
@@ -44,8 +44,9 @@ class ActivityRegularizer(Regularizer):
         self.layer = layer
 
     def __call__(self, loss):
-        loss += self.l1 * T.sum(T.mean(abs(self.layer.get_output(True)), axis=0))
-        loss += self.l2 * T.sum(T.mean(self.layer.get_output(True) ** 2, axis=0))
+        output = self.layer.get_output(True)
+        loss += self.l1 * K.sum(K.mean(K.abs(output), axis=0))
+        loss += self.l2 * K.sum(K.mean(K.square(output), axis=0))
         return loss
 
     def get_config(self):
@@ -81,4 +82,5 @@ identity = Regularizer
 
 from .utils.generic_utils import get_from_module
 def get(identifier, kwargs=None):
-    return get_from_module(identifier, globals(), 'regularizer', instantiate=True, kwargs=kwargs)
+    return get_from_module(identifier, globals(), 'regularizer',
+                           instantiate=True, kwargs=kwargs)

keras/utils/generic_utils.py

@@ -107,7 +107,10 @@ class Progbar(object):
             else:
                 info += ' - %ds' % (now - self.start)
             for k in self.unique_values:
-                info += ' - %s: %.4f' % (k, self.sum_values[k][0] / max(1, self.sum_values[k][1]))
+                if type(self.sum_values[k]) is list:
+                    info += ' - %s: %.4f' % (k, self.sum_values[k][0] / max(1, self.sum_values[k][1]))
+                else:
+                    info += ' - %s: %s' % (k, self.sum_values[k])
 
             self.total_width += len(info)
             if prev_total_width > self.total_width:

keras/utils/layer_utils.py

@@ -1,26 +1,29 @@
 from __future__ import print_function
 import inspect
 import numpy as np
-import theano
 import copy
 
-from ..layers.advanced_activations import LeakyReLU, PReLU
-from ..layers.core import Dense, Merge, Dropout, Activation, Reshape, Flatten, RepeatVector, Layer, AutoEncoder
-from ..layers.core import ActivityRegularization, TimeDistributedDense, AutoEncoder, MaxoutDense
-from ..layers.convolutional import Convolution1D, Convolution2D, MaxPooling1D, MaxPooling2D, ZeroPadding2D
-from ..layers.embeddings import Embedding, WordContextProduct
-from ..layers.noise import GaussianNoise, GaussianDropout
-from ..layers.normalization import BatchNormalization
-from ..layers.recurrent import SimpleRNN, SimpleDeepRNN, GRU, LSTM, JZS1, JZS2, JZS3
+from ..layers.advanced_activations import *
+from ..layers.core import *
+from ..layers.convolutional import *
+from ..layers.embeddings import *
+from ..layers.noise import *
+from ..layers.normalization import *
+from ..layers.recurrent import *
 from ..layers import containers
 from .. import regularizers
 from .. import constraints
 
 
-def container_from_config(original_layer_dict):
+def container_from_config(original_layer_dict, custom_objects={}):
     layer_dict = copy.deepcopy(original_layer_dict)
     name = layer_dict.get('name')
 
+    # Insert custom layers into globals so they can
+    # be accessed by `get_from_module`.
+    for cls_key in custom_objects:
+        globals()[cls_key] = custom_objects[cls_key]
+
     if name == 'Merge':
         mode = layer_dict.get('mode')
         layers = layer_dict.get('layers')
@@ -70,56 +73,21 @@ def container_from_config(original_layer_dict):
         layer_dict.pop('name')
 
         for k, v in layer_dict.items():
-            # For now, this can only happen for regularizers and constraints
             if isinstance(v, dict):
-                vname = v.get('name')
-                v.pop('name')
+                vname = v.pop('name')
                 if vname in [x for x, y in inspect.getmembers(constraints, predicate=inspect.isclass)]:
                     layer_dict[k] = constraints.get(vname, v)
-                if vname in [x for x, y in inspect.getmembers(regularizers, predicate=inspect.isclass)]:
+                elif vname in [x for x, y in inspect.getmembers(regularizers, predicate=inspect.isclass)]:
                     layer_dict[k] = regularizers.get(vname, v)
+                else:
+                    # not a regularizer of constraint, don't touch it
+                    v['name'] = vname
 
         base_layer = get_layer(name, layer_dict)
         return base_layer
 
 
-def print_layer_shapes(model, input_shapes):
-    """
-    Utility function to print the shape of the output at each layer of a Model
-
-    Arguments:
-        model: instance of Model / Merge
-        input_shapes: dict (Graph), list of tuples (Merge) or tuple (Sequential)
-    """
-    if model.__class__.__name__ in ['Sequential', 'Merge']:
-        # in this case input_shapes is a tuple, or a list [shape1, shape2]
-        if not isinstance(input_shapes[0], tuple):
-            input_shapes = [input_shapes]
-
-        inputs = model.get_input(train=False)
-        if not isinstance(inputs, list):
-            inputs = [inputs]
-        input_dummy = [np.zeros(shape, dtype=np.float32)
-                       for shape in input_shapes]
-        layers = model.layers
-
-    elif model.__class__.__name__ == 'Graph':
-        # in this case input_shapes is a dictionary
-        inputs = [model.inputs[name].input
-                  for name in model.input_order]
-        input_dummy = [np.zeros(input_shapes[name], dtype=np.float32)
-                       for name in model.input_order]
-        layers = [model.nodes[c['name']] for c in model.node_config]
-
-    print("input shapes : ", input_shapes)
-    for l in layers:
-        shape_f = theano.function(inputs, l.get_output(train=False).shape,
-                                  on_unused_input='ignore')
-        out_shape = tuple(shape_f(*input_dummy))
-        config = l.get_config()
-        print('shape after %s: %s' % (config['name'], out_shape))
-
-
 from .generic_utils import get_from_module
 def get_layer(identifier, kwargs=None):
-    return get_from_module(identifier, globals(), 'layer', instantiate=True, kwargs=kwargs)
+    return get_from_module(identifier, globals(), 'layer',
+                           instantiate=True, kwargs=kwargs)

keras/utils/np_utils.py

@@ -34,7 +34,6 @@ def binary_logloss(p, y):
 
 
 def multiclass_logloss(P, Y):
-    score = 0.
     npreds = [P[i][Y[i]-1] for i in range(len(Y))]
     score = -(1. / len(Y)) * np.sum(np.log(npreds))
     return score

keras/utils/theano_utils.py

@@ -1,40 +0,0 @@
-from __future__ import absolute_import
-import numpy as np
-import theano
-import theano.tensor as T
-
-
-def floatX(X):
-    return np.asarray(X, dtype=theano.config.floatX)
-
-
-def sharedX(X, dtype=theano.config.floatX, name=None):
-    return theano.shared(np.asarray(X, dtype=dtype), name=name)
-
-
-def shared_zeros(shape, dtype=theano.config.floatX, name=None):
-    return sharedX(np.zeros(shape), dtype=dtype, name=name)
-
-
-def shared_scalar(val=0., dtype=theano.config.floatX, name=None):
-    return theano.shared(np.cast[dtype](val))
-
-
-def shared_ones(shape, dtype=theano.config.floatX, name=None):
-    return sharedX(np.ones(shape), dtype=dtype, name=name)
-
-
-def alloc_zeros_matrix(*dims):
-    return T.alloc(np.cast[theano.config.floatX](0.), *dims)
-
-
-def ndim_tensor(ndim):
-    if ndim == 1:
-        return T.vector()
-    elif ndim == 2:
-        return T.matrix()
-    elif ndim == 3:
-        return T.tensor3()
-    elif ndim == 4:
-        return T.tensor4()
-    return T.matrix()

keras/utils/visualize_util.py

@@ -0,0 +1,41 @@
+import pydot
+# old pydot will not work with python3, must use one
+# that works with python3 such as pydot2 or pydot
+from keras.models import Sequential, Graph
+
+def to_graph(model):
+    graph = pydot.Dot(graph_type='digraph')
+    if type(model) == Sequential:
+        previous_node = None
+        written_nodes = []
+        n = 1
+        for node in model.get_config()['layers']:
+            # append number in case layers have same name to differentiate
+            if (node['name'] + str(n)) in written_nodes:
+                n += 1
+            current_node = pydot.Node(node['name'] + str(n))
+            written_nodes.append(node['name'] + str(n))
+            graph.add_node(current_node)
+            if previous_node:
+                graph.add_edge(pydot.Edge(previous_node, current_node))
+            previous_node = current_node
+    elif type(model) == Graph:
+        # don't need to append number for names since all nodes labeled
+        for input_node in model.input_config:
+            graph.add_node(pydot.Node(input_node['name']))
+
+        # intermediate and output nodes have input defined
+        for layer_config in [model.node_config, model.output_config]:
+            for node in layer_config:
+                graph.add_node(pydot.Node(node['name']))
+                # possible to have multiple 'inputs' vs 1 'input'
+                if node['inputs']:
+                    for e in node['inputs']:
+                        graph.add_edge(pydot.Edge(e, node['name']))
+                else:
+                    graph.add_edge(pydot.Edge(node['input'], node['name']))
+    return graph
+
+def plot(model, to_file='model.png'):
+    graph = to_graph(model)
+    graph.write_png(to_file)

setup.py

@@ -3,12 +3,15 @@ from setuptools import find_packages
 
 
 setup(name='Keras',
-      version='0.1.2',
+      version='0.2.0',
       description='Theano-based Deep Learning library',
       author='Francois Chollet',
       author_email='francois.chollet@gmail.com',
       url='https://github.com/fchollet/keras',
-      download_url='https://github.com/fchollet/keras/tarball/0.1.2',
+      download_url='https://github.com/fchollet/keras/tarball/0.2.0',
       license='MIT',
-      install_requires=['theano', 'pyyaml', 'h5py'],
+      install_requires=['theano', 'pyyaml', 'six'],
+      extras_require={
+          'h5py': ['h5py'],
+      },
       packages=find_packages())

tests/auto/keras/layers/test_recurrent.py

@@ -1,59 +0,0 @@
-import unittest
-import numpy as np
-import theano
-
-from keras.layers import recurrent
-
-nb_samples, timesteps, input_dim, output_dim = 3, 3, 10, 5
-
-
-def _runner(layer_class):
-    """
-    All the recurrent layers share the same interface, so we can run through them with a single
-    function.
-    """
-    for weights in [None, [np.ones((input_dim, output_dim))]]:
-        for ret_seq in [True, False]:
-            layer = layer_class(input_dim, output_dim, return_sequences=ret_seq, weights=weights)
-            layer.input = theano.shared(value=np.ones((nb_samples, timesteps, input_dim)))
-            config = layer.get_config()
-
-            for train in [True, False]:
-                out = layer.get_output(train).eval()
-                # Make sure the output has the desired shape
-                if ret_seq:
-                    assert(out.shape == (nb_samples, timesteps, output_dim))
-                else:
-                    assert(out.shape == (nb_samples, output_dim))
-
-                mask = layer.get_output_mask(train)
-
-
-class TestRNNS(unittest.TestCase):
-    """
-    Test all the RNNs using a generic test runner function defined above.
-    """
-    def test_simple(self):
-        _runner(recurrent.SimpleRNN)
-
-    def test_simple_deep(self):
-        _runner(recurrent.SimpleDeepRNN)
-
-    def test_gru(self):
-        _runner(recurrent.GRU)
-
-    def test_lstm(self):
-        _runner(recurrent.LSTM)
-
-    def test_jzs1(self):
-        _runner(recurrent.JZS1)
-
-    def test_jzs2(self):
-        _runner(recurrent.JZS2)
-
-    def test_jzs3(self):
-        _runner(recurrent.JZS3)
-
-
-if __name__ == '__main__':
-    unittest.main()

tests/auto/keras/test_activations.py

@@ -1,104 +0,0 @@
-import math
-
-import keras
-import theano
-import theano.tensor as T
-
-import numpy
-
-def list_assert_equal(a, b, round_to=7):
-    '''
-    This will do a pairwise, rounded equality test across two lists of
-    numbers.
-    '''
-    pairs = zip(a, b)
-    for i, j in pairs:
-        assert round(i, round_to) == round(j, round_to)
-
-def get_standard_values():
-    '''
-    These are just a set of floats used for testing the activation
-    functions, and are useful in multiple tests.
-    '''
-
-    return [0,0.1,0.5,0.9,1.0]
-
-def test_softmax():
-
-    from keras.activations import softmax as s
-
-    # Test using a reference implementation of softmax
-    def softmax(values):
-        m = max(values)
-        values = numpy.array(values)
-        e = numpy.exp(values - m)
-        dist = list(e / numpy.sum(e))
-
-        return dist
-
-    x = T.vector()
-    exp = s(x)
-    f = theano.function([x], exp)
-    test_values=get_standard_values()
-
-    result = f(test_values)
-    expected = softmax(test_values)
-
-    print(str(result))
-    print(str(expected))
-
-    list_assert_equal(result, expected)
-
-def test_relu():
-    '''
-    Relu implementation doesn't depend on the value being
-    a theano variable. Testing ints, floats and theano tensors.
-    '''
-
-    from keras.activations import relu as r
-
-    assert r(5) == 5
-    assert r(-5) == 0
-    assert r(-0.1) == 0
-    assert r(0.1) == 0.1
-
-    x = T.vector()
-    exp = r(x)
-    f = theano.function([x], exp)
-
-    test_values = get_standard_values()
-    result = f(test_values)
-
-    list_assert_equal(result, test_values) # because no negatives in test values
-
-
-def test_tanh():
-
-    from keras.activations import tanh as t
-    test_values = get_standard_values()
-
-    x = T.vector()
-    exp = t(x)
-    f = theano.function([x], exp)
-
-    result = f(test_values)
-    expected = [math.tanh(v) for v in test_values]
-
-    print(result)
-    print(expected)
-
-    list_assert_equal(result, expected)
-
-
-def test_linear():
-    '''
-    This function does no input validation, it just returns the thing
-    that was passed in.
-    '''
-
-    from keras.activations import linear as l
-
-    xs = [1, 5, True, None, 'foo']
-
-    for x in xs:
-        assert x == l(x)

tests/auto/keras/test_normalization.py

@@ -1,99 +0,0 @@
-import unittest
-import numpy as np
-from numpy.testing import assert_allclose
-from theano import tensor as T
-from keras.layers import normalization
-from keras.models import Sequential
-
-
-class TestBatchNormalization(unittest.TestCase):
-    def setUp(self):
-        self.input_1 = np.arange(10)
-        self.input_2 = np.zeros(10)
-        self.input_3 = np.ones((10))
-
-        self.input_shapes = [np.ones((10, 10)), np.ones((10, 10, 10))]
-
-    def test_setup(self):
-        norm_m0 = normalization.BatchNormalization((10, 10))
-        norm_m1 = normalization.BatchNormalization((10, 10), mode=1)
-
-        # mode 3 does not exist
-        self.assertRaises(Exception, normalization.BatchNormalization((10, 10), mode=3))
-
-    def test_mode_0(self):
-        model = Sequential()
-        norm_m0 = normalization.BatchNormalization((10,))
-        model.add(norm_m0)
-        model.compile(loss='mse', optimizer='sgd')
-
-        # centered on 5.0, variance 10.0
-        X = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10))
-        model.fit(X, X, nb_epoch=5, verbose=0)
-        norm_m0.input = X
-        out = (norm_m0.get_output(train=True) - norm_m0.beta) / norm_m0.gamma
-
-        self.assertAlmostEqual(out.mean().eval(), 0.0, places=1)
-        self.assertAlmostEqual(out.std().eval(), 1.0, places=1)
-
-    def test_mode_1(self):
-        norm_m1 = normalization.BatchNormalization((10,), mode=1)
-        norm_m1.init_updates()
-
-        for inp in [self.input_1, self.input_2, self.input_3]:
-            norm_m1.input = inp
-            out = (norm_m1.get_output(train=True) - norm_m1.beta) / norm_m1.gamma
-            self.assertAlmostEqual(out.mean().eval(), 0.0)
-            if inp.std() > 0.:
-                self.assertAlmostEqual(out.std().eval(), 1.0, places=2)
-            else:
-                self.assertAlmostEqual(out.std().eval(), 0.0, places=2)
-
-    def test_shapes(self):
-        """
-        Test batch normalization with various input shapes
-        """
-        for inp in self.input_shapes:
-            norm_m0 = normalization.BatchNormalization(inp.shape, mode=0)
-            norm_m0.init_updates()
-            norm_m0.input = inp
-            out = (norm_m0.get_output(train=True) - norm_m0.beta) / norm_m0.gamma
-
-            norm_m1 = normalization.BatchNormalization(inp.shape, mode=1)
-            norm_m1.input = inp
-            out = (norm_m1.get_output(train=True) - norm_m1.beta) / norm_m1.gamma
-
-    def test_weight_init(self):
-        """
-        Test weight initialization
-        """
-
-        norm_m1 = normalization.BatchNormalization((10,), mode=1, weights=[np.ones(10), np.ones(10)])
-        norm_m1.init_updates()
-
-        for inp in [self.input_1, self.input_2, self.input_3]:
-            norm_m1.input = inp
-            out = (norm_m1.get_output(train=True) - np.ones(10)) / 1.
-            self.assertAlmostEqual(out.mean().eval(), 0.0)
-            if inp.std() > 0.:
-                self.assertAlmostEqual(out.std().eval(), 1.0, places=2)
-            else:
-                self.assertAlmostEqual(out.std().eval(), 0.0, places=2)
-
-        assert_allclose(norm_m1.gamma.eval(), np.ones(10))
-        assert_allclose(norm_m1.beta.eval(), np.ones(10))
-
-        # Weights must be an iterable of gamma AND beta.
-        self.assertRaises(Exception, normalization.BatchNormalization((10,)), weights=np.ones(10))
-
-    def test_config(self):
-        norm = normalization.BatchNormalization((10, 10), mode=1, epsilon=0.1)
-        conf = norm.get_config()
-        conf_target = {"input_shape": (10, 10), "name": normalization.BatchNormalization.__name__,
-                       "epsilon": 0.1, "mode": 1}
-
-        self.assertDictEqual(conf, conf_target)
-
-
-if __name__ == '__main__':
-    unittest.main()

tests/auto/test_loss_masking.py

@@ -1,46 +0,0 @@
-import numpy as np
-np.random.seed(1337)
-
-import unittest
-from keras.models import Sequential, weighted_objective
-from keras.layers.core import TimeDistributedDense, Masking
-from keras import objectives
-import theano
-
-
-class TestLossMasking(unittest.TestCase):
-    def test_loss_masking(self):
-        X = np.array(
-            [[[1, 1], [2, 1], [3, 1], [5, 5]],
-             [[1, 5], [5, 0], [0, 0], [0, 0]]], dtype=np.int32)
-        model = Sequential()
-        model.add(Masking(mask_value=0))
-        model.add(TimeDistributedDense(2, 1, init='one'))
-        model.compile(loss='mse', optimizer='sgd')
-        y = model.predict(X)
-        loss = model.fit(X, 4*y, nb_epoch=1, batch_size=2, verbose=1).history['loss'][0]
-        assert loss == 285.
-
-    def test_loss_masking_time(self):
-        theano.config.mode = 'FAST_COMPILE'
-        weighted_loss = weighted_objective(objectives.get('categorical_crossentropy'))
-        shape = (3, 4, 2)
-        X = np.arange(24).reshape(shape)
-        Y = 2 * X
-
-        weights = np.ones((3, 4, 1))  # Normally the trailing 1 is added by standardize_weights
-        weights[0, 0] = 0
-        mask = np.ones((3, 4))
-        mask[1, 0] = 0
-
-        out = weighted_loss(X, Y, weights, mask).eval()
-        weights[0, 0] = 1e-9  # so that nonzero() doesn't remove this weight
-        out2 = weighted_loss(X, Y, weights, mask).eval()
-        print(out)
-        print(out2)
-        assert abs(out - out2) < 1e-8
-
-
-if __name__ == '__main__':
-    print('Test loss masking')
-    unittest.main()

tests/keras/backend/test_backends.py

@@ -0,0 +1,335 @@
+import sys
+import unittest
+from numpy.testing import assert_allclose
+import numpy as np
+import pytest
+
+if sys.version_info.major == 2:
+    from keras.backend import theano_backend as KTH
+    from keras.backend import tensorflow_backend as KTF
+
+
+def check_single_tensor_operation(function_name, input_shape, **kwargs):
+    val = np.random.random(input_shape) - 0.5
+    xth = KTH.variable(val)
+    xtf = KTF.variable(val)
+
+    zth = KTH.eval(getattr(KTH, function_name)(xth, **kwargs))
+    ztf = KTF.eval(getattr(KTF, function_name)(xtf, **kwargs))
+
+    assert zth.shape == ztf.shape
+    assert_allclose(zth, ztf, atol=1e-06)
+
+
+def check_two_tensor_operation(function_name, x_input_shape,
+                               y_input_shape, **kwargs):
+    xval = np.random.random(x_input_shape) - 0.5
+    xth = KTH.variable(xval)
+    xtf = KTF.variable(xval)
+
+    yval = np.random.random(y_input_shape) - 0.5
+    yth = KTH.variable(yval)
+    ytf = KTF.variable(yval)
+
+    zth = KTH.eval(getattr(KTH, function_name)(xth, yth, **kwargs))
+    ztf = KTF.eval(getattr(KTF, function_name)(xtf, ytf, **kwargs))
+
+    assert zth.shape == ztf.shape
+    assert_allclose(zth, ztf, atol=1e-06)
+
+
+@pytest.mark.skipif(sys.version_info.major != 2, reason="Requires Python 2.7")
+class TestBackend(unittest.TestCase):
+
+    def test_linear_operations(self):
+        check_two_tensor_operation('dot', (4, 2), (2, 4))
+        check_single_tensor_operation('transpose', (4, 2))
+
+    def test_shape_operations(self):
+        # concatenate
+        xval = np.random.random((4, 3))
+        xth = KTH.variable(xval)
+        xtf = KTF.variable(xval)
+        yval = np.random.random((4, 2))
+        yth = KTH.variable(yval)
+        ytf = KTF.variable(yval)
+        zth = KTH.eval(KTH.concatenate([xth, yth], axis=-1))
+        ztf = KTF.eval(KTF.concatenate([xtf, ytf], axis=-1))
+        assert zth.shape == ztf.shape
+        assert_allclose(zth, ztf, atol=1e-06)
+
+        check_single_tensor_operation('reshape', (4, 2), shape=(8, 1))
+        check_single_tensor_operation('permute_dimensions', (4, 2, 3),
+                                      pattern=(2, 0, 1))
+        check_single_tensor_operation('repeat', (4, 1), n=3)
+        check_single_tensor_operation('flatten', (4, 1))
+        check_single_tensor_operation('expand_dims', (4, 3), dim=-1)
+        check_single_tensor_operation('expand_dims', (4, 3, 2), dim=1)
+        check_single_tensor_operation('squeeze', (4, 3, 1), axis=2)
+
+    def test_value_manipulation(self):
+        val = np.random.random((4, 2))
+        xth = KTH.variable(val)
+        xtf = KTF.variable(val)
+
+        # get_value
+        valth = KTH.get_value(xth)
+        valtf = KTF.get_value(xtf)
+        assert valtf.shape == valth.shape
+        assert_allclose(valth, valtf, atol=1e-06)
+
+        # set_value
+        val = np.random.random((4, 2))
+        KTH.set_value(xth, val)
+        KTF.set_value(xtf, val)
+
+        valth = KTH.get_value(xth)
+        valtf = KTF.get_value(xtf)
+        assert valtf.shape == valth.shape
+        assert_allclose(valth, valtf, atol=1e-06)
+
+        # count_params
+        assert KTH.count_params(xth) == KTF.count_params(xtf)
+
+    def test_elementwise_operations(self):
+        check_single_tensor_operation('max', (4, 2))
+        check_single_tensor_operation('max', (4, 2), axis=1, keepdims=True)
+
+        check_single_tensor_operation('min', (4, 2))
+        check_single_tensor_operation('min', (4, 2), axis=1, keepdims=True)
+
+        check_single_tensor_operation('mean', (4, 2))
+        check_single_tensor_operation('mean', (4, 2), axis=1, keepdims=True)
+        check_single_tensor_operation('mean', (4, 2, 3), axis=-1, keepdims=True)
+
+        check_single_tensor_operation('std', (4, 2))
+        check_single_tensor_operation('std', (4, 2), axis=1, keepdims=True)
+
+        check_single_tensor_operation('prod', (4, 2))
+        check_single_tensor_operation('prod', (4, 2), axis=1, keepdims=True)
+
+        # does not work yet, wait for bool <-> int casting in TF (coming soon)
+        # check_single_tensor_operation('any', (4, 2))
+        # check_single_tensor_operation('any', (4, 2), axis=1, keepdims=True)
+
+        check_single_tensor_operation('argmax', (4, 2))
+        check_single_tensor_operation('argmax', (4, 2), axis=1)
+
+        check_single_tensor_operation('argmin', (4, 2))
+        check_single_tensor_operation('argmin', (4, 2), axis=1)
+
+        check_single_tensor_operation('square', (4, 2))
+        check_single_tensor_operation('abs', (4, 2))
+        check_single_tensor_operation('sqrt', (4, 2))
+        check_single_tensor_operation('exp', (4, 2))
+        check_single_tensor_operation('log', (4, 2))
+        check_single_tensor_operation('round', (4, 2))
+        check_single_tensor_operation('pow', (4, 2), a=3)
+        check_single_tensor_operation('clip', (4, 2), min_value=0.4,
+                                      max_value=0.6)
+
+        # two-tensor ops
+        check_two_tensor_operation('equal', (4, 2), (4, 2))
+        check_two_tensor_operation('maximum', (4, 2), (4, 2))
+        check_two_tensor_operation('minimum', (4, 2), (4, 2))
+
+    def test_gradient(self):
+        val = np.random.random((4, 2))
+        xth = KTH.variable(val)
+        xtf = KTF.variable(val)
+
+        expth = xth * KTH.exp(xth)
+        exptf = xtf * KTF.exp(xtf)
+        lossth = KTH.sum(expth)
+        losstf = KTF.sum(exptf)
+
+        gradth = KTH.gradients(lossth, [expth])
+        gradtf = KTF.gradients(losstf, [exptf])
+
+        zth = KTH.eval(gradth[0])
+        ztf = KTF.eval(gradtf[0])
+        assert zth.shape == ztf.shape
+        assert_allclose(zth, ztf, atol=1e-06)
+
+    def test_function(self):
+        val = np.random.random((4, 2))
+        input_val = np.random.random((4, 2))
+
+        xth = KTH.variable(val)
+        xtf = KTF.variable(val)
+        yth = KTH.placeholder(ndim=2)
+        ytf = KTF.placeholder(ndim=2)
+
+        exp_th = KTH.square(xth) + yth
+        exp_tf = KTF.square(xtf) + ytf
+
+        update_th = xth * 2
+        update_tf = xtf * 2
+        fth = KTH.function([yth], [exp_th], updates=[(xth, update_th)])
+        ftf = KTF.function([ytf], [exp_tf], updates=[(xtf, update_tf)])
+
+        function_outputs_th = fth([input_val])[0]
+        function_outputs_tf = ftf([input_val])[0]
+        assert function_outputs_th.shape == function_outputs_tf.shape
+        assert_allclose(function_outputs_th, function_outputs_tf, atol=1e-06)
+
+        new_val_th = KTH.get_value(xth)
+        new_val_tf = KTF.get_value(xtf)
+        assert new_val_th.shape == new_val_tf.shape
+        assert_allclose(new_val_th, new_val_tf, atol=1e-06)
+
+    def test_rnn(self):
+        # implement a simple RNN
+        input_dim = 8
+        output_dim = 4
+        timesteps = 5
+
+        input_val = np.random.random((32, timesteps, input_dim))
+        init_state_val = np.random.random((32, output_dim))
+        W_i_val = np.random.random((input_dim, output_dim))
+        W_o_val = np.random.random((output_dim, output_dim))
+
+        def rnn_step_fn(input_dim, output_dim, K):
+            W_i = K.variable(W_i_val)
+            W_o = K.variable(W_o_val)
+
+            def step_function(x, states):
+                assert len(states) == 1
+                prev_output = states[0]
+                output = K.dot(x, W_i) + K.dot(prev_output, W_o)
+                return output, [output]
+            return step_function
+
+        th_rnn_step_fn = rnn_step_fn(input_dim, output_dim, KTH)
+        inputs = KTH.variable(input_val)
+        initial_states = [KTH.variable(init_state_val)]
+        last_output, outputs, new_states = KTH.rnn(th_rnn_step_fn, inputs,
+                                                   initial_states,
+                                                   go_backwards=False,
+                                                   masking=False)
+        th_last_output = KTH.eval(last_output)
+        th_outputs = KTH.eval(outputs)
+        assert len(new_states) == 1
+        th_state = KTH.eval(new_states[0])
+
+        tf_rnn_step_fn = rnn_step_fn(input_dim, output_dim, KTF)
+        inputs = KTF.variable(input_val)
+        initial_states = [KTF.variable(init_state_val)]
+        last_output, outputs, new_states = KTF.rnn(tf_rnn_step_fn, inputs,
+                                                   initial_states,
+                                                   go_backwards=False,
+                                                   masking=False)
+        tf_last_output = KTF.eval(last_output)
+        tf_outputs = KTF.eval(outputs)
+        assert len(new_states) == 1
+        tf_state = KTF.eval(new_states[0])
+
+        assert_allclose(tf_last_output, th_last_output, atol=1e-06)
+        assert_allclose(tf_outputs, th_outputs, atol=1e-06)
+        assert_allclose(tf_state, th_state, atol=1e-06)
+
+    def test_switch(self):
+        val = np.random.random()
+        xth = KTH.variable(val)
+        xth = KTH.switch(xth >= 0.5, xth * 0.1, xth * 0.2)
+
+        xtf = KTF.variable(val)
+        xtf = KTF.switch(xtf >= 0.5, xtf * 0.1, xtf * 0.2)
+
+        zth = KTH.eval(xth)
+        ztf = KTF.eval(xtf)
+
+        assert zth.shape == ztf.shape
+        assert_allclose(zth, ztf, atol=1e-06)
+
+    def test_nn_operations(self):
+        check_single_tensor_operation('relu', (4, 2), alpha=0.1, max_value=0.5)
+        check_single_tensor_operation('softmax', (4, 10))
+        check_single_tensor_operation('softplus', (4, 10))
+
+        check_single_tensor_operation('sigmoid', (4, 2))
+        check_single_tensor_operation('hard_sigmoid', (4, 2))
+        check_single_tensor_operation('tanh', (4, 2))
+
+        # dropout
+        val = np.random.random((20, 20))
+        xth = KTH.variable(val)
+        xtf = KTF.variable(val)
+        zth = KTH.eval(KTH.dropout(xth, level=0.2))
+        ztf = KTF.eval(KTF.dropout(xtf, level=0.2))
+        assert zth.shape == ztf.shape
+        # dropout patterns are different, only check mean
+        assert np.abs(zth.mean() - ztf.mean()) < 0.05
+
+        check_two_tensor_operation('binary_crossentropy', (4, 2), (4, 2), from_logits=True)
+        check_two_tensor_operation('categorical_crossentropy', (4, 2), (4, 2), from_logits=True)
+        check_two_tensor_operation('binary_crossentropy', (4, 2), (4, 2), from_logits=False)
+
+        check_two_tensor_operation('categorical_crossentropy', (4, 2), (4, 2), from_logits=False)
+
+    # def test_conv2d(self):
+    #     '''conv2d works "properly" with Theano and TF but outputs different
+    #     values in each case. Cause unclear (input / kernel shape format?)
+    #     '''
+    #     # TH kernel shape: (depth, input_depth, rows, cols)
+    #     check_two_tensor_operation('conv2d', (5, 3, 10, 12), (4, 3, 2, 2),
+    #                                strides=(1, 1), border_mode='valid')
+    #     check_two_tensor_operation('conv2d', (5, 3, 10, 12), (4, 3, 2, 2),
+    #                                strides=(1, 1), border_mode='same')
+
+    #     # TF kernel shape: (rows, cols, input_depth, depth)
+    #     check_two_tensor_operation('conv2d', (5, 10, 12, 3), (2, 2, 3, 4),
+    #                                strides=(1, 1), border_mode='valid', dim_ordering='tf')
+    #     check_two_tensor_operation('conv2d', (5, 10, 12, 3), (2, 2, 3, 4),
+    #                                strides=(1, 1), border_mode='same', dim_ordering='tf')
+
+    #     check_two_tensor_operation('conv2d', (5, 3, 10, 12), (4, 3, 3, 3),
+    #                                strides=(1, 1), border_mode='valid')
+    #     check_two_tensor_operation('conv2d', (5, 3, 10, 12), (4, 3, 3, 3),
+    #                                strides=(1, 1), border_mode='same')
+
+    #     check_two_tensor_operation('conv2d', (5, 3, 10, 12), (4, 3, 3, 3),
+    #                                strides=(2, 2), border_mode='valid')
+
+    # def test_maxpool2d(self):
+    #     '''maxpool2d works "properly" with Theano and TF but outputs different
+    #     values in each case. Cause unclear (input shape format?)
+    #     '''
+    #     check_single_tensor_operation('maxpool2d', (5, 3, 10, 12), pool_size=(2, 2),
+    #                                   strides=(1, 1), border_mode='valid')
+
+    #     check_single_tensor_operation('maxpool2d', (5, 3, 9, 11), pool_size=(2, 2),
+    #                                   strides=(1, 1), border_mode='valid')
+
+    #     check_single_tensor_operation('maxpool2d', (5, 3, 9, 11), pool_size=(2, 3),
+    #                                   strides=(1, 1), border_mode='valid')
+
+    def test_random_normal(self):
+        mean = 0.
+        std = 1.
+        rand = KTF.get_value(KTF.random_normal((1000, 1000), mean=mean, std=std))
+        assert(rand.shape == (1000, 1000))
+        assert(np.abs(np.mean(rand) - mean) < 0.01)
+        assert(np.abs(np.std(rand) - std) < 0.01)
+
+        rand = KTF.get_value(KTF.random_normal((1000, 1000), mean=mean, std=std))
+        assert(rand.shape == (1000, 1000))
+        assert(np.abs(np.mean(rand) - mean) < 0.01)
+        assert(np.abs(np.std(rand) - std) < 0.01)
+
+    def test_random_uniform(self):
+        mean = 0.
+        std = 1.
+        rand = KTF.get_value(KTF.random_normal((1000, 1000), mean=mean, std=std))
+        assert(rand.shape == (1000, 1000))
+        assert(np.abs(np.mean(rand) - mean) < 0.01)
+        assert(np.abs(np.std(rand) - std) < 0.01)
+
+        rand = KTF.get_value(KTF.random_normal((1000, 1000), mean=mean, std=std))
+        assert(rand.shape == (1000, 1000))
+        assert(np.abs(np.mean(rand) - mean) < 0.01)
+        assert(np.abs(np.std(rand) - std) < 0.01)
+
+
+if __name__ == '__main__':
+    unittest.main()

tests/keras/layers/test_convolutional.py

@@ -0,0 +1,166 @@
+import unittest
+import numpy as np
+from numpy.testing import assert_allclose
+
+from keras import backend as K
+from keras.layers import convolutional
+
+
+class TestConvolutions(unittest.TestCase):
+    def test_convolution_1d(self):
+        nb_samples = 9
+        nb_steps = 7
+        input_dim = 10
+        filter_length = 6
+        nb_filter = 5
+
+        weights_in = [np.ones((nb_filter, input_dim, filter_length, 1)),
+                      np.ones(nb_filter)]
+
+        input = np.ones((nb_samples, nb_steps, input_dim))
+        for weight in [None, weights_in]:
+            for border_mode in ['valid', 'same']:
+                for subsample_length in [1]:
+                    if border_mode == 'same' and subsample_length != 1:
+                        continue
+                    for W_regularizer in [None, 'l2']:
+                        for b_regularizer in [None, 'l2']:
+                            for act_regularizer in [None, 'l2']:
+                                layer = convolutional.Convolution1D(
+                                    nb_filter, filter_length,
+                                    weights=weight,
+                                    border_mode=border_mode,
+                                    W_regularizer=W_regularizer,
+                                    b_regularizer=b_regularizer,
+                                    activity_regularizer=act_regularizer,
+                                    subsample_length=subsample_length,
+                                    input_shape=(None, input_dim))
+
+                            layer.input = K.variable(input)
+                            for train in [True, False]:
+                                out = K.eval(layer.get_output(train))
+                                assert input.shape[0] == out.shape[0]
+                                if border_mode == 'same' and subsample_length == 1:
+                                    assert input.shape[1] == out.shape[1]
+                            layer.get_config()
+
+    def test_maxpooling_1d(self):
+        nb_samples = 9
+        nb_steps = 7
+        input_dim = 10
+
+        input = np.ones((nb_samples, nb_steps, input_dim))
+        for stride in [1, 2]:
+            layer = convolutional.MaxPooling1D(stride=stride,
+                                               border_mode='valid')
+            layer.input = K.variable(input)
+            for train in [True, False]:
+                K.eval(layer.get_output(train))
+            layer.get_config()
+
+    def test_convolution_2d(self):
+        nb_samples = 8
+        nb_filter = 9
+        stack_size = 7
+        nb_row = 10
+        nb_col = 6
+
+        input_nb_row = 11
+        input_nb_col = 12
+
+        weights_in = [np.ones((nb_filter, stack_size, nb_row, nb_col)), np.ones(nb_filter)]
+
+        input = np.ones((nb_samples, stack_size, input_nb_row, input_nb_col))
+        for weight in [None, weights_in]:
+            for border_mode in ['valid', 'same']:
+                for subsample in [(1, 1), (2, 2)]:
+                    if border_mode == 'same' and subsample != (1, 1):
+                        continue
+                    for W_regularizer in [None, 'l2']:
+                        for b_regularizer in [None, 'l2']:
+                            for act_regularizer in [None, 'l2']:
+                                layer = convolutional.Convolution2D(
+                                    nb_filter, nb_row, nb_col,
+                                    weights=weight,
+                                    border_mode=border_mode,
+                                    W_regularizer=W_regularizer,
+                                    b_regularizer=b_regularizer,
+                                    activity_regularizer=act_regularizer,
+                                    subsample=subsample,
+                                    input_shape=(stack_size, None, None))
+
+                                layer.input = K.variable(input)
+                                for train in [True, False]:
+                                    out = K.eval(layer.get_output(train))
+                                    if border_mode == 'same' and subsample == (1, 1):
+                                        assert out.shape[2:] == input.shape[2:]
+                                layer.get_config()
+
+    def test_maxpooling_2d(self):
+        nb_samples = 9
+        stack_size = 7
+        input_nb_row = 11
+        input_nb_col = 12
+        pool_size = (3, 3)
+
+        input = np.ones((nb_samples, stack_size, input_nb_row, input_nb_col))
+        for strides in [(1, 1), (2, 2)]:
+            layer = convolutional.MaxPooling2D(strides=strides,
+                                               border_mode='valid',
+                                               pool_size=pool_size)
+            layer.input = K.variable(input)
+            for train in [True, False]:
+                K.eval(layer.get_output(train))
+            layer.get_config()
+
+    def test_zero_padding_2d(self):
+        nb_samples = 9
+        stack_size = 7
+        input_nb_row = 11
+        input_nb_col = 12
+
+        input = np.ones((nb_samples, stack_size, input_nb_row, input_nb_col))
+        layer = convolutional.ZeroPadding2D(padding=(2, 2))
+        layer.input = K.variable(input)
+        for train in [True, False]:
+            out = K.eval(layer.get_output(train))
+            for offset in [0, 1, -1, -2]:
+                assert_allclose(out[:, :, offset, :], 0.)
+                assert_allclose(out[:, :, :, offset], 0.)
+            assert_allclose(out[:, :, 2:-2, 2:-2], 1.)
+        layer.get_config()
+
+    def test_upsampling_1d(self):
+        nb_samples = 9
+        nb_steps = 7
+        input_dim = 10
+
+        input = np.ones((nb_samples, nb_steps, input_dim))
+        for length in [2, 3, 9]:
+            layer = convolutional.UpSampling1D(length=length)
+            layer.input = K.variable(input)
+            for train in [True, False]:
+                out = K.eval(layer.get_output(train))
+                assert out.shape[1] == length * nb_steps
+            layer.get_config()
+
+    def test_upsampling_2d(self):
+        nb_samples = 9
+        stack_size = 7
+        input_nb_row = 11
+        input_nb_col = 12
+
+        input = np.ones((nb_samples, stack_size, input_nb_row, input_nb_col))
+
+        for length_row in [2, 3, 9]:
+            for length_col in [2, 3, 9]:
+                layer = convolutional.UpSampling2D(size=(length_row, length_col))
+                layer.input = K.variable(input)
+                for train in [True, False]:
+                    out = K.eval(layer.get_output(train))
+                    assert out.shape[2] == length_row * input_nb_row
+                    assert out.shape[3] == length_col * input_nb_col
+            layer.get_config()
+
+if __name__ == '__main__':
+    unittest.main()

tests/keras/layers/test_core.py

@@ -1,8 +1,8 @@
 import unittest
 import numpy as np
 from numpy.testing import assert_allclose
-import theano
 
+from keras import backend as K
 from keras.layers import core
 
 
@@ -12,18 +12,13 @@ class TestLayerBase(unittest.TestCase):
         input_dim = 5
         layer = core.Layer()
 
-        # As long as there is no input, an error should be raised.
-        for train in [True, False]:
-            self.assertRaises(AttributeError, layer.get_input, train)
-            self.assertRaises(AttributeError, layer.get_output, train)
-
         # Once an input is provided, it should be reachable through the
         # appropriate getters
         input = np.ones((nb_samples, input_dim))
-        layer.input = theano.shared(value=input)
+        layer.input = K.variable(input)
         for train in [True, False]:
-            assert_allclose(layer.get_input(train).eval(), input)
-            assert_allclose(layer.get_output(train).eval(), input)
+            assert_allclose(K.eval(layer.get_input(train)), input)
+            assert_allclose(K.eval(layer.get_output(train)), input)
 
     def test_connections(self):
         nb_samples = 10
@@ -32,17 +27,13 @@ class TestLayerBase(unittest.TestCase):
         layer2 = core.Layer()
 
         input = np.ones((nb_samples, input_dim))
-        layer1.input = theano.shared(value=input)
-
-        # As long as there is no previous layer, an error should be raised.
-        for train in [True, False]:
-            self.assertRaises(AttributeError, layer2.get_input, train)
+        layer1.input = K.variable(input)
 
         # After connecting, input of layer1 should be passed through
         layer2.set_previous(layer1)
         for train in [True, False]:
-            assert_allclose(layer2.get_input(train).eval(), input)
-            assert_allclose(layer2.get_output(train).eval(), input)
+            assert_allclose(K.eval(layer2.get_input(train)), input)
+            assert_allclose(K.eval(layer2.get_output(train)), input)
 
 
 class TestConfigParams(unittest.TestCase):
@@ -69,6 +60,8 @@ class TestConfigParams(unittest.TestCase):
     def test_merge(self):
         layer_1 = core.Layer()
         layer_2 = core.Layer()
+        layer_1.set_input_shape((None,))
+        layer_2.set_input_shape((None,))
         layer = core.Merge([layer_1, layer_2])
         self._runner(layer)
 
@@ -81,7 +74,7 @@ class TestConfigParams(unittest.TestCase):
         self._runner(layer)
 
     def test_reshape(self):
-        layer = core.Reshape(10, 10)
+        layer = core.Reshape(dims=(10, 10))
         self._runner(layer)
 
     def test_flatten(self):
@@ -93,7 +86,7 @@ class TestConfigParams(unittest.TestCase):
         self._runner(layer)
 
     def test_dense(self):
-        layer = core.Dense(10, 10)
+        layer = core.Dense(10, input_shape=(10,))
         self._runner(layer)
 
     def test_act_reg(self):
@@ -101,7 +94,11 @@ class TestConfigParams(unittest.TestCase):
         self._runner(layer)
 
     def test_time_dist_dense(self):
-        layer = core.TimeDistributedDense(10, 10)
+        layer = core.TimeDistributedDense(10, input_shape=(None, 10))
+        self._runner(layer)
+
+    def test_time_dist_merge(self):
+        layer = core.TimeDistributedMerge()
         self._runner(layer)
 
     def test_autoencoder(self):
@@ -121,41 +118,48 @@ class TestMasking(unittest.TestCase):
 
     def test_sequences(self):
         """Test masking sequences with zeroes as padding"""
+        if K._BACKEND == "tensorflow":
+            return
         # integer inputs, one per timestep, like embeddings
         layer = core.Masking()
-        func = theano.function([layer.input], layer.get_output_mask())
-        self.assertTrue(np.all(
-            # get mask for this input
-            func(np.array(
-            [[[1], [2], [3], [0]],
-             [[0], [4], [5], [0]]], dtype=np.int32)) ==
-            # This is the expected output mask, one dimension less
-            np.array([[1, 1, 1, 0], [0, 1, 1, 0]])))
+        func = K.function([layer.input], [layer.get_output_mask()])
+        input_data = np.array([[[1], [2], [3], [0]],
+                              [[0], [4], [5], [0]]], dtype=np.int32)
+
+        # This is the expected output mask, one dimension less
+        expected = np.array([[1, 1, 1, 0], [0, 1, 1, 0]])
+
+        # get mask for this input
+        output = func([input_data])[0]
+        self.assertTrue(np.all(output == expected))
 
     def test_non_zero(self):
         """Test masking with non-zero mask value"""
+        if K._BACKEND == "tensorflow":
+            return
         layer = core.Masking(5)
-        func = theano.function([layer.input], layer.get_output_mask())
-        self.assertTrue(np.all(
-            # get mask for this input, if not all the values are 5, shouldn't masked
-            func(np.array(
-            [[[1, 1], [2, 1], [3, 1], [5, 5]],
-             [[1, 5], [5, 0], [0, 0], [0, 0]]], dtype=np.int32)) ==
-            # This is the expected output mask, one dimension less
-            np.array([[1, 1, 1, 0], [1, 1, 1, 1]])))
+        func = K.function([layer.input], [layer.get_output_mask()])
+        input_data = np.array([[[1, 1], [2, 1], [3, 1], [5, 5]],
+                              [[1, 5], [5, 0], [0, 0], [0, 0]]],
+                              dtype=np.int32)
+        output = func([input_data])[0]
+        expected = np.array([[1, 1, 1, 0], [1, 1, 1, 1]])
+        self.assertTrue(np.all(output == expected))
 
     def test_non_zero_output(self):
         """Test output of masking layer with non-zero mask value"""
+        if K._BACKEND == "tensorflow":
+            return
         layer = core.Masking(5)
-        func = theano.function([layer.input], layer.get_output())
-        self.assertTrue(np.all(
-            # get output for this input, replace padding with 0
-            func(np.array(
-            [[[1, 1], [2, 1], [3, 1], [5, 5]],
-             [[1, 5], [5, 0], [0, 0], [0, 0]]], dtype=np.int32)) ==
-            # This is the expected output
-            np.array([[[1, 1], [2, 1], [3, 1], [0, 0]],
-             [[1, 5], [5, 0], [0, 0], [0, 0]]])))
+        func = K.function([layer.input], [layer.get_output()])
+
+        input_data = np.array([[[1, 1], [2, 1], [3, 1], [5, 5]],
+                              [[1, 5], [5, 0], [0, 0], [0, 0]]],
+                              dtype=np.int32)
+        output = func([input_data])[0]
+        expected = np.array([[[1, 1], [2, 1], [3, 1], [0, 0]],
+                            [[1, 5], [5, 0], [0, 0], [0, 0]]])
+        self.assertTrue(np.all(output == expected))
 
 
 if __name__ == '__main__':

tests/keras/layers/test_recurrent.py

@@ -0,0 +1,47 @@
+import unittest
+import numpy as np
+
+from keras.layers import recurrent
+from keras import backend as K
+
+nb_samples, timesteps, input_dim, output_dim = 3, 3, 10, 5
+
+
+def _runner(layer_class):
+    """
+    All the recurrent layers share the same interface,
+    so we can run through them with a single function.
+    """
+    for ret_seq in [True, False]:
+        layer = layer_class(output_dim, return_sequences=ret_seq,
+                            weights=None, input_shape=(timesteps, input_dim))
+        layer.input = K.variable(np.ones((nb_samples, timesteps, input_dim)))
+        layer.get_config()
+
+        for train in [True, False]:
+            out = K.eval(layer.get_output(train))
+            # Make sure the output has the desired shape
+            if ret_seq:
+                assert(out.shape == (nb_samples, timesteps, output_dim))
+            else:
+                assert(out.shape == (nb_samples, output_dim))
+
+            mask = layer.get_output_mask(train)
+
+
+class TestRNNS(unittest.TestCase):
+    """
+    Test all the RNNs using a generic test runner function defined above.
+    """
+    def test_simple(self):
+        _runner(recurrent.SimpleRNN)
+
+    def test_gru(self):
+        _runner(recurrent.GRU)
+
+    def test_lstm(self):
+        _runner(recurrent.LSTM)
+
+
+if __name__ == '__main__':
+    unittest.main()

tests/keras/test_activations.py

@@ -0,0 +1,76 @@
+import unittest
+from keras import backend as K
+import numpy as np
+from numpy.testing import assert_allclose
+
+
+def get_standard_values():
+    '''
+    These are just a set of floats used for testing the activation
+    functions, and are useful in multiple tests.
+    '''
+    return np.array([[0, 0.1, 0.5, 0.9, 1.0]], dtype=K.floatx())
+
+
+class TestActivations(unittest.TestCase):
+
+    def test_softmax(self):
+        from keras.activations import softmax as s
+
+        # Test using a reference implementation of softmax
+        def softmax(values):
+            m = np.max(values)
+            e = np.exp(values - m)
+            return e / np.sum(e)
+
+        x = K.placeholder(ndim=2)
+        exp = s(x)
+        f = K.function([x], [exp])
+        test_values = get_standard_values()
+
+        result = f([test_values])[0]
+        expected = softmax(test_values)
+        assert_allclose(result, expected, rtol=1e-05)
+
+    def test_relu(self):
+        '''
+        Relu implementation doesn't depend on the value being
+        a theano variable. Testing ints, floats and theano tensors.
+        '''
+        from keras.activations import relu as r
+
+        x = K.placeholder(ndim=2)
+        exp = r(x)
+        f = K.function([x], [exp])
+
+        test_values = get_standard_values()
+        result = f([test_values])[0]
+
+        # because no negatives in test values
+        assert_allclose(result, test_values, rtol=1e-05)
+
+    def test_tanh(self):
+        from keras.activations import tanh as t
+        test_values = get_standard_values()
+
+        x = K.placeholder(ndim=2)
+        exp = t(x)
+        f = K.function([x], [exp])
+
+        result = f([test_values])[0]
+        expected = np.tanh(test_values)
+        assert_allclose(result, expected, rtol=1e-05)
+
+    def test_linear(self):
+        '''
+        This function does no input validation, it just returns the thing
+        that was passed in.
+        '''
+        from keras.activations import linear as l
+
+        xs = [1, 5, True, None, 'foo']
+        for x in xs:
+            assert x == l(x)
+
+if __name__ == '__main__':
+    unittest.main()

tests/keras/test_constraints.py

@@ -1,7 +1,7 @@
 import unittest
 import numpy as np
 from numpy.testing import assert_allclose
-from theano import tensor as T
+from keras import backend as K
 
 
 class TestConstraints(unittest.TestCase):
@@ -16,23 +16,25 @@ class TestConstraints(unittest.TestCase):
 
         for m in self.some_values:
             norm_instance = maxnorm(m)
-            normed = norm_instance(self.example_array)
-            assert (np.all(normed.eval() < m))
+            normed = norm_instance(K.variable(self.example_array))
+            assert (np.all(K.eval(normed) < m))
 
         # a more explicit example
         norm_instance = maxnorm(2.0)
         x = np.array([[0, 0, 0], [1.0, 0, 0], [3, 0, 0], [3, 3, 3]]).T
-        x_normed_target = np.array([[0, 0, 0], [1.0, 0, 0], [2.0, 0, 0], [2./np.sqrt(3), 2./np.sqrt(3), 2./np.sqrt(3)]]).T
-        x_normed_actual = norm_instance(x).eval()
-        assert_allclose(x_normed_actual, x_normed_target)
+        x_normed_target = np.array([[0, 0, 0], [1.0, 0, 0],
+                                    [2.0, 0, 0],
+                                    [2. / np.sqrt(3), 2. / np.sqrt(3), 2. / np.sqrt(3)]]).T
+        x_normed_actual = K.eval(norm_instance(K.variable(x)))
+        assert_allclose(x_normed_actual, x_normed_target, rtol=1e-05)
 
     def test_nonneg(self):
         from keras.constraints import nonneg
 
         nonneg_instance = nonneg()
 
-        normed = nonneg_instance(self.example_array)
-        assert (np.all(np.min(normed.eval(), axis=1) == 0.))
+        normed = nonneg_instance(K.variable(self.example_array))
+        assert (np.all(np.min(K.eval(normed), axis=1) == 0.))
 
     def test_identity(self):
         from keras.constraints import identity
@@ -58,12 +60,13 @@ class TestConstraints(unittest.TestCase):
         from keras.constraints import unitnorm
         unitnorm_instance = unitnorm()
 
-        normalized = unitnorm_instance(self.example_array)
+        normalized = unitnorm_instance(K.variable(self.example_array))
 
-        norm_of_normalized = np.sqrt(np.sum(normalized.eval()**2, axis=1))
-        difference = norm_of_normalized - 1. #in the unit norm constraint, it should be equal to 1.
+        norm_of_normalized = np.sqrt(np.sum(K.eval(normalized)**2, axis=1))
+        # in the unit norm constraint, it should be equal to 1.
+        difference = norm_of_normalized - 1.
         largest_difference = np.max(np.abs(difference))
-        self.assertAlmostEqual(largest_difference, 0.)
+        assert np.abs(largest_difference) < 10e-5
 
 if __name__ == '__main__':
     unittest.main()

tests/keras/test_normalization.py

@@ -0,0 +1,102 @@
+import unittest
+import numpy as np
+from numpy.testing import assert_allclose
+from keras.layers import normalization
+from keras.models import Sequential
+from keras import backend as K
+
+
+class TestBatchNormalization(unittest.TestCase):
+    def setUp(self):
+        self.input_1 = np.arange(10)
+        self.input_2 = np.zeros(10)
+        self.input_3 = np.ones((10))
+
+        self.input_shapes = [np.ones((10, 10)), np.ones((10, 10, 10))]
+
+    def test_setup(self):
+        norm_m0 = normalization.BatchNormalization(input_shape=(10, 10))
+        norm_m1 = normalization.BatchNormalization(input_shape=(10, 10), mode=1)
+
+        # mode 3 does not exist
+        self.assertRaises(Exception,
+                          normalization.BatchNormalization(input_shape=(10, 10), mode=3))
+
+    def test_mode_0(self):
+        model = Sequential()
+        norm_m0 = normalization.BatchNormalization(input_shape=(10,))
+        model.add(norm_m0)
+        model.compile(loss='mse', optimizer='sgd')
+
+        # centered on 5.0, variance 10.0
+        X = np.random.normal(loc=5.0, scale=10.0, size=(1000, 10))
+        model.fit(X, X, nb_epoch=5, verbose=0)
+        norm_m0.input = K.variable(X)
+        out = (norm_m0.get_output(train=True) - norm_m0.beta) / norm_m0.gamma
+
+        self.assertAlmostEqual(K.eval(K.mean(out)), 0.0, places=1)
+        self.assertAlmostEqual(K.eval(K.std(out)), 1.0, places=1)
+
+    def test_mode_1(self):
+        norm_m1 = normalization.BatchNormalization(input_shape=(10,), mode=1)
+
+        for inp in [self.input_1, self.input_2, self.input_3]:
+            norm_m1.input = K.variable(inp)
+            out = (norm_m1.get_output(train=True) - norm_m1.beta) / norm_m1.gamma
+            self.assertAlmostEqual(K.eval(K.mean(out)), 0.0)
+            if inp.std() > 0.:
+                self.assertAlmostEqual(K.eval(K.std(out)), 1.0, places=2)
+            else:
+                self.assertAlmostEqual(K.eval(K.std(out)), 0.0, places=2)
+
+    def test_shapes(self):
+        """
+        Test batch normalization with various input shapes
+        """
+        for inp in self.input_shapes:
+            norm_m0 = normalization.BatchNormalization(input_shape=inp.shape, mode=0)
+            norm_m0.input = K.variable(inp)
+            out = (norm_m0.get_output(train=True) - norm_m0.beta) / norm_m0.gamma
+
+            norm_m1 = normalization.BatchNormalization(input_shape=inp.shape, mode=1)
+            norm_m1.input = K.variable(inp)
+            out = (norm_m1.get_output(train=True) - norm_m1.beta) / norm_m1.gamma
+
+    def test_weight_init(self):
+        """
+        Test weight initialization
+        """
+        norm_m1 = normalization.BatchNormalization(input_shape=(10,), mode=1,
+                                                   weights=[np.ones(10), np.ones(10), np.zeros(10), np.zeros(10)])
+
+        for inp in [self.input_1, self.input_2, self.input_3]:
+            norm_m1.input = K.variable(inp)
+            out = (norm_m1.get_output(train=True) - np.ones(10)) / 1.
+            self.assertAlmostEqual(K.eval(K.mean(out)), 0.0)
+            if inp.std() > 0.:
+                self.assertAlmostEqual(K.eval(K.std(out)), 1.0, places=2)
+            else:
+                self.assertAlmostEqual(K.eval(K.std(out)), 0.0, places=2)
+
+        assert_allclose(K.eval(norm_m1.gamma), np.ones(10))
+        assert_allclose(K.eval(norm_m1.beta), np.ones(10))
+
+    def test_config(self):
+        norm = normalization.BatchNormalization(input_shape=(10, 10), mode=1,
+                                                epsilon=0.1, momentum=0.9)
+        conf = norm.get_config()
+        conf_target = {"input_shape": (10, 10),
+                       "name": normalization.BatchNormalization.__name__,
+                       "epsilon": 0.1, "mode": 1, "momentum": 0.9}
+        self.assertDictEqual(conf, conf_target)
+
+    def test_save_weights(self):
+        norm = normalization.BatchNormalization(input_shape=(10, 10), mode=1,
+                                                epsilon=0.1)
+        weights = norm.get_weights()
+        assert(len(weights) == 4)
+        norm.set_weights(weights)
+
+
+if __name__ == '__main__':
+    unittest.main()

tests/manual/check_autoencoder.py

@@ -1,134 +0,0 @@
-from __future__ import absolute_import
-from __future__ import print_function
-from keras.datasets import mnist
-from keras.models import Sequential, model_from_config
-from keras.layers.core import AutoEncoder, Dense, Activation, TimeDistributedDense, Flatten
-from keras.layers.recurrent import LSTM
-from keras.layers.embeddings import Embedding
-from keras.layers.core import Layer
-from keras.layers import containers
-from keras.utils import np_utils
-import numpy as np
-
-nb_classes = 10
-batch_size = 128
-nb_epoch = 5
-activation = 'linear'
-
-input_dim = 784
-hidden_dim = 392
-
-max_train_samples = 5000
-max_test_samples = 1000
-
-# the data, shuffled and split between tran and test sets
-(X_train, y_train), (X_test, y_test) = mnist.load_data()
-
-X_train = X_train.reshape(60000, input_dim)[:max_train_samples]
-X_test = X_test.reshape(10000, input_dim)[:max_test_samples]
-X_train = X_train.astype("float32")
-X_test = X_test.astype("float32")
-X_train /= 255
-X_test /= 255
-
-# convert class vectors to binary class matrices
-Y_train = np_utils.to_categorical(y_train, nb_classes)[:max_train_samples]
-Y_test = np_utils.to_categorical(y_test, nb_classes)[:max_test_samples]
-
-print("X_train: ", X_train.shape)
-print("X_test: ", X_test.shape)
-
-
-##########################
-# dense model test       #
-##########################
-
-print("Training classical fully connected layer for classification")
-model_classical = Sequential()
-model_classical.add(Dense(input_dim, 10, activation=activation))
-model_classical.add(Activation('softmax'))
-model_classical.get_config(verbose=1)
-model_classical.compile(loss='categorical_crossentropy', optimizer='adam')
-model_classical.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=0, validation_data=(X_test, Y_test))
-classical_score = model_classical.evaluate(X_test, Y_test, verbose=0, show_accuracy=True)
-print('\nclassical_score:', classical_score)
-
-##########################
-# autoencoder model test #
-##########################
-
-
-def build_lstm_autoencoder(autoencoder, X_train, X_test):
-    X_train = X_train[:, np.newaxis, :]
-    X_test = X_test[:, np.newaxis, :]
-    print("Modified X_train: ", X_train.shape)
-    print("Modified X_test: ", X_test.shape)
-
-    # The TimeDistributedDense isn't really necessary, however you need a lot of GPU memory to do 784x394-394x784
-    autoencoder.add(TimeDistributedDense(input_dim, 16))
-    autoencoder.add(AutoEncoder(encoder=LSTM(16, 8, activation=activation, return_sequences=True),
-                                decoder=LSTM(8, input_dim, activation=activation, return_sequences=True),
-                                output_reconstruction=False))
-    return autoencoder, X_train, X_test
-
-
-def build_deep_classical_autoencoder(autoencoder):
-    encoder = containers.Sequential([Dense(input_dim, hidden_dim, activation=activation), Dense(hidden_dim, hidden_dim/2, activation=activation)])
-    decoder = containers.Sequential([Dense(hidden_dim/2, hidden_dim, activation=activation), Dense(hidden_dim, input_dim, activation=activation)])
-    autoencoder.add(AutoEncoder(encoder=encoder, decoder=decoder, output_reconstruction=False))
-    return autoencoder
-
-# Try different things here: 'lstm' or 'classical' or 'denoising'
-# or 'deep_denoising'
-
-for autoencoder_type in ['classical', 'lstm']:
-    print(autoencoder_type)
-    print('-'*40)
-    # Build our autoencoder model
-    autoencoder = Sequential()
-    if autoencoder_type == 'lstm':
-        print("Training LSTM AutoEncoder")
-        autoencoder, X_train, X_test = build_lstm_autoencoder(autoencoder, X_train, X_test)
-    elif autoencoder_type == 'classical':
-        print("Training Classical AutoEncoder")
-        autoencoder = build_deep_classical_autoencoder(autoencoder)
-    else:
-        print("Error: unknown autoencoder type!")
-        exit(-1)
-
-    autoencoder.compile(loss='mean_squared_error', optimizer='adam')
-    # Do NOT use validation data with return output_reconstruction=True
-    autoencoder.fit(X_train, X_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=1)
-
-    # Do an inference pass
-    prefilter_train = autoencoder.predict(X_train, verbose=0)
-    prefilter_test = autoencoder.predict(X_test, verbose=0)
-    print("prefilter_train: ", prefilter_train.shape)
-    print("prefilter_test: ", prefilter_test.shape)
-
-    # Classify results from Autoencoder
-    print("Building classical fully connected layer for classification")
-    model = Sequential()
-    if autoencoder_type == 'lstm':
-        model.add(TimeDistributedDense(8, nb_classes, activation=activation))
-        model.add(Flatten())
-    elif autoencoder_type == 'classical':
-        model.add(Dense(prefilter_train.shape[1], nb_classes, activation=activation))
-    else:
-        model.add(Dense(prefilter_train.shape[1], nb_classes, activation=activation))
-
-    model.add(Activation('softmax'))
-
-    model.get_config(verbose=1)
-    model.compile(loss='categorical_crossentropy', optimizer='adam')
-    model.fit(prefilter_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=0, validation_data=(prefilter_test, Y_test))
-
-    score = model.evaluate(prefilter_test, Y_test, verbose=0, show_accuracy=True)
-    print('\nscore:', score)
-
-    print('Loss change:', (score[0] - classical_score[0])/classical_score[0], '%')
-    print('Accuracy change:', (score[1] - classical_score[1])/classical_score[1], '%')
-
-    # check serialization
-    config = autoencoder.get_config(verbose=1)
-    autoencoder = model_from_config(config)

tests/manual/check_callbacks.py

@@ -1,226 +0,0 @@
-import numpy as np
-import random
-import theano
-
-from keras.models import Sequential
-from keras.callbacks import Callback
-from keras.layers.core import Dense, Dropout, Activation, Flatten
-from keras.regularizers import l2
-from keras.layers.convolutional import Convolution2D, MaxPooling2D
-from keras.utils import np_utils
-from keras.datasets import mnist
-import keras.callbacks as cbks
-
-from matplotlib import pyplot as plt
-from matplotlib import animation
-
-##############################
-# model DrawActivations test #
-##############################
-
-print('Running DrawActivations test')
-
-nb_classes = 10
-batch_size = 128
-nb_epoch = 10
-
-max_train_samples = 512
-max_test_samples = 1
-
-np.random.seed(1337)
-
-# the data, shuffled and split between tran and test sets
-(X_train, y_train), (X_test, y_test) = mnist.load_data()
-
-X_train = X_train.reshape(-1,1,28,28)[:max_train_samples]
-X_train = X_train.astype("float32")
-X_train /= 255
-
-X_test = X_test.reshape(-1,1,28,28)[:max_test_samples]
-X_test = X_test.astype("float32")
-X_test /= 255
-
-# convert class vectors to binary class matrices
-Y_train = np_utils.to_categorical(y_train, nb_classes)[:max_train_samples]
-
-class Frames(object):
-    def __init__(self, n_plots=16):
-        self._n_frames = 0
-        self._framedata = []
-        self._titles = []
-        for i in range(n_plots):
-            self._framedata.append([])
-
-    def add_frame(self, i, frame):
-        self._framedata[i].append(frame)
-
-    def set_title(self, title):
-        self._titles.append(title)
-
-class SubplotTimedAnimation(animation.TimedAnimation):
-
-    def __init__(self, fig, frames, grid=(4, 4), interval=10, blit=False, **kwargs):
-        self.n_plots = grid[0] * grid[1]
-        self.axes = [fig.add_subplot(grid[0], grid[1], i + 1) for i in range(self.n_plots)]
-        for axis in self.axes:
-            axis.get_xaxis().set_ticks([])
-            axis.get_yaxis().set_ticks([])
-        self.frames = frames
-        self.imgs = [self.axes[i].imshow(frames._framedata[i][0], interpolation='nearest', cmap='bone') for i in range(self.n_plots)]
-        self.title = fig.suptitle('')
-        super(SubplotTimedAnimation, self).__init__(fig, interval=interval, blit=blit, **kwargs)
-
-    def _draw_frame(self, j):
-        for i in range(self.n_plots):
-            self.imgs[i].set_data(self.frames._framedata[i][j])
-        if len(self.frames._titles) > j:
-            self.title.set_text(self.frames._titles[j])
-        self._drawn_artists = self.imgs
-
-    def new_frame_seq(self):
-        return iter(range(len(self.frames._framedata[0])))
-
-    def _init_draw(self):
-        for img in self.imgs:
-            img.set_data([[]])
-
-def combine_imgs(imgs, grid=(1,1)):
-    n_imgs, img_h, img_w = imgs.shape
-    if n_imgs != grid[0] * grid[1]:
-        raise ValueError()
-    combined = np.zeros((grid[0] * img_h, grid[1] * img_w))
-    for i in range(grid[0]):
-        for j in range(grid[1]):
-            combined[img_h*i:img_h*(i+1),img_w*j:img_w*(j+1)] = imgs[grid[0] * i + j]
-    return combined
-
-class DrawActivations(Callback):
-    def __init__(self, figsize):
-        self.fig = plt.figure(figsize=figsize)
-
-    def on_train_begin(self, logs={}):
-        self.imgs = Frames(n_plots=5)
-
-        layers_0_ids = np.random.choice(32, 16, replace=False)
-        self.test_layer0 = theano.function([self.model.get_input()], self.model.layers[1].get_output(train=False)[0, layers_0_ids])
-
-        layers_1_ids = np.random.choice(64, 36, replace=False)
-        self.test_layer1 = theano.function([self.model.get_input()], self.model.layers[5].get_output(train=False)[0, layers_1_ids])
-
-        self.test_layer2 = theano.function([self.model.get_input()], self.model.layers[10].get_output(train=False)[0])
-
-    def on_epoch_begin(self, epoch, logs={}):
-        self.epoch = epoch
-
-    def on_batch_end(self, batch, logs={}):
-        if batch % 5 == 0:
-            self.imgs.add_frame(0, X_test[0,0])
-            self.imgs.add_frame(1, combine_imgs(self.test_layer0(X_test), grid=(4, 4)))
-            self.imgs.add_frame(2, combine_imgs(self.test_layer1(X_test), grid=(6, 6)))
-            self.imgs.add_frame(3, self.test_layer2(X_test).reshape((16,16)))
-            self.imgs.add_frame(4, self.model._predict(X_test)[0].reshape((1,10)))
-            self.imgs.set_title('Epoch #%d - Batch #%d' % (self.epoch, batch))
-
-    def on_train_end(self, logs={}):
-        anim = SubplotTimedAnimation(self.fig, self.imgs, grid=(1,5), interval=10, blit=False, repeat_delay=1000)
-        # anim.save('test_gif.gif', fps=15, writer='imagemagick')
-        plt.show()
-
-# model = Sequential()
-# model.add(Dense(784, 50))
-# model.add(Activation('relu'))
-# model.add(Dense(50, 10))
-# model.add(Activation('softmax'))
-
-model = Sequential()
-model.add(Convolution2D(32, 1, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-model.add(Dropout(0.25))
-
-model.add(Convolution2D(64, 32, 3, 3, border_mode='full')) 
-model.add(Activation('relu'))
-model.add(MaxPooling2D(poolsize=(2, 2)))
-model.add(Dropout(0.25))
-
-model.add(Flatten())
-model.add(Dense(64*8*8, 256))
-model.add(Activation('relu'))
-model.add(Dropout(0.5))
-
-model.add(Dense(256, 10, W_regularizer = l2(0.1)))
-model.add(Activation('softmax'))
-
-model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
-
-# Fit the model
-draw_weights = DrawActivations(figsize=(5.4, 1.35))
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, verbose=1, callbacks=[draw_weights])
-
-
-##########################
-# model checkpoint tests #
-##########################
-
-print('Running ModelCheckpoint test')
-
-nb_classes = 10
-batch_size = 128
-nb_epoch = 20
-
-# small sample size to overfit on training data
-max_train_samples = 50
-max_test_samples = 1000
-
-np.random.seed(1337) # for reproducibility
-
-# the data, shuffled and split between tran and test sets
-(X_train, y_train), (X_test, y_test) = mnist.load_data()
-
-X_train = X_train.reshape(60000,784)[:max_train_samples]
-X_test = X_test.reshape(10000,784)[:max_test_samples]
-X_train = X_train.astype("float32")
-X_test = X_test.astype("float32")
-X_train /= 255
-X_test /= 255
-
-# convert class vectors to binary class matrices
-Y_train = np_utils.to_categorical(y_train, nb_classes)[:max_train_samples]
-Y_test = np_utils.to_categorical(y_test, nb_classes)[:max_test_samples]
-
-
-# Create a slightly larger network than required to test best validation save only
-model = Sequential()
-model.add(Dense(784, 500))
-model.add(Activation('relu'))
-model.add(Dense(500, 10))
-model.add(Activation('softmax'))
-model.compile(loss='categorical_crossentropy', optimizer='rmsprop')
-
-# test file location
-path = "/tmp"
-filename = "model_weights.hdf5"
-import os
-f = os.path.join(path, filename)
-
-print("Test model checkpointer")
-# only store best validation model in checkpointer
-checkpointer = cbks.ModelCheckpoint(filepath=f, verbose=1, save_best_only=True)
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=0, validation_data=(X_test, Y_test), callbacks =[checkpointer])
-
-if not os.path.isfile(f):
-    raise Exception("Model weights were not saved to %s" % (f))
-
-print("Test model checkpointer without validation data")
-import warnings
-warnings.filterwarnings('error')
-try:
-    # this should issue a warning
-    model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=0, callbacks =[checkpointer])
-except:
-    print("Tests passed")
-    import sys
-    sys.exit(0)
-
-raise Exception("Modelcheckpoint tests did not pass")
-

tests/manual/check_constraints.py

@@ -1,100 +0,0 @@
-from __future__ import absolute_import
-from __future__ import print_function
-import keras
-from keras.datasets import mnist
-import keras.models
-from keras.models import Sequential
-from keras.layers.core import Dense, Dropout, Activation
-from keras.regularizers import l2, l1
-from keras.constraints import maxnorm, nonneg
-from keras.optimizers import SGD, Adam, RMSprop
-from keras.utils import np_utils, generic_utils
-import theano
-import theano.tensor as T
-import numpy as np
-import scipy
-
-batch_size = 100
-nb_classes = 10
-nb_epoch = 10
-
-# the data, shuffled and split between tran and test sets
-(X_train, y_train), (X_test, y_test) = mnist.load_data()
-X_train=X_train.reshape(60000,784)
-X_test=X_test.reshape(10000,784)
-X_train = X_train.astype("float32")
-X_test = X_test.astype("float32")
-X_train /= 255
-X_test /= 255
-
-# convert class vectors to binary class matrices
-Y_train = np_utils.to_categorical(y_train, nb_classes)
-Y_test = np_utils.to_categorical(y_test, nb_classes)
-
-model = Sequential()
-model.add(Dense(784, 20, W_constraint=maxnorm(1)))
-model.add(Activation('relu'))
-model.add(Dropout(0.1))
-model.add(Dense(20, 20, W_constraint=nonneg()))
-model.add(Activation('relu'))
-model.add(Dropout(0.1))
-model.add(Dense(20, 10, W_constraint=maxnorm(1)))
-model.add(Activation('softmax'))
-
-
-rms = RMSprop()
-model.compile(loss='categorical_crossentropy', optimizer=rms)
-
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=True, verbose=0)
-
-a=model.params[0].eval()
-if np.isclose(np.max(np.sqrt(np.sum(a**2, axis=0))),1):
-	print('Maxnorm test passed')
-else:
-	raise ValueError('Maxnorm test failed!')
-		
-b=model.params[2].eval()
-if np.min(b)==0 and np.min(a)!=0:
-	print('Nonneg test passed')
-else:
-	raise ValueError('Nonneg test failed!')
-	
-
-model = Sequential()
-model.add(Dense(784, 20))
-model.add(Activation('relu'))
-model.add(Dense(20, 20, W_regularizer=l1(.01)))
-model.add(Activation('relu'))
-model.add(Dense(20, 10))
-model.add(Activation('softmax'))
-
-
-rms = RMSprop()
-model.compile(loss='categorical_crossentropy', optimizer=rms)
-
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=20, show_accuracy=True, verbose=0)
-
-a=model.params[2].eval().reshape(400)
-(D, p1) = scipy.stats.kurtosistest(a)
-
-model = Sequential()
-model.add(Dense(784, 20))
-model.add(Activation('relu'))
-model.add(Dense(20, 20, W_regularizer=l2(.01)))
-model.add(Activation('relu'))
-model.add(Dense(20, 10))
-model.add(Activation('softmax'))
-
-
-rms = RMSprop()
-model.compile(loss='categorical_crossentropy', optimizer=rms)
-
-model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=20, show_accuracy=True, verbose=0)
-
-a=model.params[2].eval().reshape(400)
-(D, p2) = scipy.stats.kurtosistest(a)
-
-if p1<.01 and p2>.01:
-	print('L1 and L2 regularization tests passed')
-else:
-	raise ValueError('L1 and L2 regularization tests failed!')

tests/manual/check_masked_recurrent.py

@@ -1,130 +0,0 @@
-# Dummy test data as input to RNN. This input is 3 timesteps long where the third timestep always matches the
-# first. Without masking it should be able to learn it, with masking it should fail.
-
-import numpy as np
-from keras.utils.theano_utils import sharedX
-from keras.models import Sequential
-from keras.layers.core import Dense, Activation, Merge, Dropout, TimeDistributedDense
-from keras.layers.embeddings import Embedding
-from keras.layers.recurrent import SimpleRNN, SimpleDeepRNN, LSTM, GRU
-import theano
-
-theano.config.exception_verbosity = 'high' 
-
-# (nb_samples, timesteps, dimensions)
-X = np.random.random_integers(1, 4, size=(500000, 15))
-
-print("About to compile the first model")
-model = Sequential()
-model.add(Embedding(5, 4, mask_zero=True))
-model.add(TimeDistributedDense(4, 4)) # obviously this is redundant. Just testing.
-model.add(SimpleRNN(4, 4, activation='relu', return_sequences=True))
-model.add(Dropout(0.5))
-model.add(SimpleDeepRNN(4, 4, depth=2, activation='relu')) 
-model.add(Dropout(0.5))
-model.add(Dense(4, 4, activation='softmax'))
-model.compile(loss='categorical_crossentropy',
-        optimizer='rmsprop', theano_mode=theano.compile.mode.FAST_RUN)
-print("Compiled model")
-
-W = model.get_weights() # We'll save these so we can reset it later
-
-X[:, : 10] = 0
-Xmask0 = X.copy()
-Xmask0[:, 10] = 0
-
-Xmask12 = X.copy()
-Xmask12[:, 11] = 0
-Xmask12[:, 12] = 0
-
-X0_onehot = np.zeros((X.shape[0], 4))
-X1_onehot = np.zeros((X.shape[0], 4))
-for i, row in enumerate(X):
-    X0_onehot[i, row[10] - 1] = 1
-    X1_onehot[i, row[11] - 1] = 1
-
-# Uniform score: 4 options = ln(4) nats (2 bits)
-# we should not do better than this when we mask out the part of the input
-# that gives us the correct answer
-uniform_score = np.log(4)
-batch_size=512
-
-# Train it to guess 0th dim
-model.fit(X, X0_onehot, nb_epoch=1, batch_size=batch_size)
-score = model.evaluate(X, X0_onehot, batch_size=batch_size)
-if score > uniform_score * 0.9:
-    raise Exception('Failed to learn to copy timestep 0, score %f' % score)
-    
-
-model.set_weights(W)
-
-# Train without showing it the 0th dim to learn 1st dim
-model.fit(X[: , 1:], X1_onehot, nb_epoch=1, batch_size=batch_size)
-score = model.evaluate(X[:, 1:], X1_onehot, batch_size=batch_size)
-if score > uniform_score * 0.9:
-    raise Exception('Failed to learn to copy timestep 1, score %f' % score)
-
-model.set_weights(W)
-
-# Train to guess 0th dim when 0th dim has been masked (should fail)
-model.fit(Xmask0, X0_onehot, nb_epoch=1, batch_size=batch_size)
-score = model.evaluate(Xmask0, X0_onehot, batch_size=batch_size)
-if score < uniform_score * 0.9:
-   raise Exception('Somehow learned to copy timestep 0 despite mask, score %f' % score)
-
-model.set_weights(W)
-
-# Train to guess 1st dim when 0th dim has been masked (should succeed)
-model.fit(Xmask0, X1_onehot, nb_epoch=1, batch_size=batch_size)
-score = model.evaluate(Xmask0, X1_onehot, batch_size=batch_size)
-if score > uniform_score * 0.9:
-    raise Exception('Failed to learn to copy timestep 1 in masked model, score %f' % score)
-
-model.set_weights(W)
-
-# Finally, make sure the mask is actually blocking input, mask out timesteps 1 and 2, and see if
-# it can learn timestep 0 (should fail)
-model.fit(Xmask12, X0_onehot, nb_epoch=1, batch_size=batch_size)
-
-score = model.evaluate(Xmask12, X0_onehot, batch_size=batch_size)
-if score < uniform_score * 0.9:
-    raise Exception('Somehow learned to copy timestep 0 despite masking 1, score %f' % score)
-
-# Another testing approach, just initialize models and make sure that prepending zeros doesn't affect
-# their output
-print("About to compile the second model")
-model2 = Sequential()
-model2.add(Embedding(5, 4, mask_zero=True))
-model2.add(TimeDistributedDense(4, 4))
-model2.add(Activation('time_distributed_softmax'))
-model2.add(LSTM(4, 4, return_sequences=True))
-model2.add(Activation('tanh'))
-model2.add(GRU(4, 4, activation='softmax', return_sequences=True))
-model2.add(SimpleDeepRNN(4, 4, depth=2, activation='relu', return_sequences=True)) 
-model2.add(SimpleRNN(4, 4, activation='relu', return_sequences=True))
-model2.compile(loss='categorical_crossentropy',
-        optimizer='rmsprop', theano_mode=theano.compile.mode.FAST_RUN)
-print("Compiled model2")
-
-X2 = np.random.random_integers(1, 4, size=(2, 5))
-y2 = np.random.random((X2.shape[0], X2.shape[1], 4))
-
-ref = model2.predict(X2)
-ref_eval = model2.evaluate(X2, y2)
-mask = np.ones((y2.shape[0], y2.shape[1], 1))
-
-for pre_zeros in range(1, 10):
-    padded_X2 = np.concatenate((np.zeros((X2.shape[0], pre_zeros)), X2), axis=1)
-    padded_mask = np.concatenate((np.zeros((mask.shape[0], pre_zeros, mask.shape[2])), mask), axis=1)
-    padded_y2 = np.concatenate((np.zeros((y2.shape[0], pre_zeros, y2.shape[2])), y2), axis=1)
-
-    pred = model2.predict(padded_X2)
-    if not np.allclose(ref[:, -1, :], pred[:, -1, :]):
-        raise Exception("Different result after left-padding %d zeros. Ref: %s, Pred: %s" % (pre_zeros, ref, pred))
-
-    pad_eval = model2.evaluate(padded_X2, padded_y2, weights=padded_mask)
-    if not np.allclose([pad_eval], [ref_eval]):
-        raise Exception("Got dissimilar categorical_crossentropy after left-padding %d zeros. Ref: %f, Pred %f" %\
-                (pref_eval, pred_val))
-
-        

tests/manual/check_model_utils.py

@@ -1,44 +0,0 @@
-from __future__ import absolute_import
-from __future__ import print_function
-from keras.models import Sequential, Graph
-from keras.layers.core import Layer, Activation, Dense, Flatten, Reshape, Merge
-from keras.layers.convolutional import Convolution2D, MaxPooling2D
-import keras.utils.layer_utils as layer_utils
-
-print('-- Sequential model')
-left = Sequential()
-left.add(Convolution2D(32, 1, 3, 3, border_mode='valid'))
-left.add(MaxPooling2D(poolsize=(2, 2)))
-left.add(Flatten())
-left.add(Dense(32 * 13 * 13, 50))
-left.add(Activation('relu'))
-
-right = Sequential()
-right.add(Dense(784, 30))
-right.add(Activation('relu'))
-
-model = Sequential()
-model.add(Merge([left, right], mode='concat'))
-
-model.add(Dense(80, 10))
-model.add(Activation('softmax'))
-
-layer_utils.print_layer_shapes(model, [(1, 1, 28, 28), (1, 784)])
-
-print('-- Graph model')
-graph = Graph()
-graph.add_input(name='input1', ndim=2)
-graph.add_input(name='input2', ndim=4)
-graph.add_node(Dense(32, 16), name='dense1', input='input1')
-graph.add_node(Dense(16, 4), name='dense3', input='dense1')
-
-graph.add_node(Convolution2D(32, 1, 3, 3), name='conv1', input='input2')
-graph.add_node(Flatten(), name='flatten1', input='conv1')
-graph.add_node(Dense(32 * 13 * 13, 10), name='dense4', input='flatten1')
-
-graph.add_output(name='output1', inputs=['dense1', 'dense3'], merge_mode='sum')
-graph.add_output(name='output2', inputs=['dense1', 'dense4'], merge_mode='concat')
-
-layer_utils.print_layer_shapes(graph, {'input1': (1, 32), 'input2': (1, 1, 28, 28)})
-
-print('Test script complete')

tests/manual/check_save_weights.py

@@ -1,40 +0,0 @@
-from keras.models import Sequential
-from keras.layers.core import Dense, Dropout, Activation
-from keras.optimizers import SGD
-
-import sys
-sys.setrecursionlimit(10000) # to be able to pickle Theano compiled functions
-
-import pickle, numpy
-
-def create_model():
-    model = Sequential()
-    model.add(Dense(256, 2048, init='uniform', activation='relu'))
-    model.add(Dropout(0.5))
-    model.add(Dense(2048, 2048, init='uniform', activation='relu'))
-    model.add(Dropout(0.5))
-    model.add(Dense(2048, 2048, init='uniform', activation='relu'))
-    model.add(Dropout(0.5))
-    model.add(Dense(2048, 2048, init='uniform', activation='relu'))
-    model.add(Dropout(0.5))
-    model.add(Dense(2048, 256, init='uniform', activation='linear'))
-    return model
-
-model = create_model()
-sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
-model.compile(loss='mse', optimizer=sgd)
-
-pickle.dump(model, open('/tmp/model.pkl', 'wb'))
-model.save_weights('/tmp/model_weights.hdf5')
-
-model_loaded = create_model()
-model_loaded.load_weights('/tmp/model_weights.hdf5')
-
-for k in range(len(model.layers)):
-    weights_orig = model.layers[k].get_weights()
-    weights_loaded = model_loaded.layers[k].get_weights()
-    for x, y in zip(weights_orig, weights_loaded):
-        if numpy.any(x != y):
-            raise ValueError('Loaded weights are different from pickled weights!')
-
-

tests/manual/check_wrappers.py

@@ -1,126 +0,0 @@
-from __future__ import absolute_import
-from __future__ import print_function
-from keras.datasets import mnist
-from keras.models import Sequential
-from keras.layers.core import Dense, Activation
-from keras.utils import np_utils
-from keras.wrappers.scikit_learn import *
-import numpy as np
-
-batch_size = 128
-nb_epoch = 1
-
-nb_classes = 10
-max_train_samples = 5000
-max_test_samples = 1000
-
-np.random.seed(1337) # for reproducibility
-
-############################################
-# scikit-learn classification wrapper test #
-############################################
-print('Beginning scikit-learn classification wrapper test')
-
-print('Loading data')
-(X_train, y_train), (X_test, y_test) = mnist.load_data()
-
-X_train = X_train.reshape(60000, 784)[:max_train_samples]
-X_test = X_test.reshape(10000, 784)[:max_test_samples]
-X_train = X_train.astype('float32')
-X_test = X_test.astype('float32')
-X_train /= 255
-X_test /= 255
-
-Y_train = np_utils.to_categorical(y_train, nb_classes)[:max_train_samples]
-Y_test = np_utils.to_categorical(y_test, nb_classes)[:max_test_samples]
-
-print('Defining model')
-model = Sequential()
-model.add(Dense(784, 50))
-model.add(Activation('relu'))
-model.add(Dense(50, 10))
-model.add(Activation('softmax'))
-
-print('Creating wrapper')
-classifier = KerasClassifier(model, train_batch_size=batch_size, nb_epoch=nb_epoch)
-
-print('Fitting model')
-classifier.fit(X_train, Y_train)
-
-print('Testing score function')
-score = classifier.score(X_train, Y_train)
-print('Score: ', score)
-
-print('Testing predict function')
-preds = classifier.predict(X_test)
-print('Preds.shape: ', preds.shape)
-
-print('Testing predict proba function')
-proba = classifier.predict_proba(X_test)
-print('Proba.shape: ', proba.shape)
-
-print('Testing get params')
-print(classifier.get_params())
-
-print('Testing set params')
-classifier.set_params(optimizer='sgd', loss='binary_crossentropy')
-print(classifier.get_params())
-
-print('Testing attributes')
-print('Classes')
-print(classifier.classes_)
-print('Config')
-print(classifier.config_)
-print('Weights')
-print(classifier.weights_)
-print('Compiled model')
-print(classifier.compiled_model_)
-
-########################################
-# scikit-learn regression wrapper test #
-########################################
-print('Beginning scikit-learn regression wrapper test')
-
-print('Generating data')
-X_train = np.random.random((5000, 100))
-X_test = np.random.random((1000, 100))
-y_train = np.random.random(5000)
-y_test = np.random.random(1000)
-
-print('Defining model')
-model = Sequential()
-model.add(Dense(100, 50))
-model.add(Activation('relu'))
-model.add(Dense(50, 1))
-model.add(Activation('linear'))
-
-print('Creating wrapper')
-regressor = KerasRegressor(model, train_batch_size=batch_size, nb_epoch=nb_epoch)
-
-print('Fitting model')
-regressor.fit(X_train, y_train)
-
-print('Testing score function')
-score = regressor.score(X_train, y_train)
-print('Score: ', score)
-
-print('Testing predict function')
-preds = regressor.predict(X_test)
-print('Preds.shape: ', preds.shape)
-
-print('Testing get params')
-print(regressor.get_params())
-
-print('Testing set params')
-regressor.set_params(optimizer='sgd', loss='mean_absolute_error')
-print(regressor.get_params())
-
-print('Testing attributes')
-print('Config')
-print(regressor.config_)
-print('Weights')
-print(regressor.weights_)
-print('Compiled model')
-print(regressor.compiled_model_)
-
-print('Test script complete.')

tests/manual/check_yaml.py

@@ -1,101 +0,0 @@
-from __future__ import absolute_import
-from __future__ import print_function
-import numpy as np
-
-from keras.utils.test_utils import get_test_data
-from keras.preprocessing import sequence
-from keras.optimizers import SGD, RMSprop, Adagrad
-from keras.utils import np_utils
-from keras.models import Sequential, Graph
-from keras.layers.core import Dense, Dropout, Activation, Merge
-from keras.layers.embeddings import Embedding
-from keras.layers.recurrent import LSTM, GRU
-from keras.datasets import imdb
-from keras.models import model_from_yaml
-
-'''
-This is essentially the IMDB test. Deserialized models should yield
-the same config as the original one.
-'''
-
-max_features = 10000
-maxlen = 100
-batch_size = 32
-
-(X_train, y_train), (X_test, y_test) = imdb.load_data(nb_words=max_features, test_split=0.2)
-
-X_train = sequence.pad_sequences(X_train, maxlen=maxlen)
-X_test = sequence.pad_sequences(X_test, maxlen=maxlen)
-
-model = Sequential()
-model.add(Embedding(max_features, 128))
-model.add(LSTM(128, 128))
-model.add(Dropout(0.5))
-model.add(Dense(128, 1, W_regularizer='identity', b_constraint='maxnorm'))
-model.add(Activation('sigmoid'))
-
-model.get_config(verbose=1)
-
-#####################################
-# save model w/o parameters to yaml #
-#####################################
-
-yaml_no_params = model.to_yaml()
-
-no_param_model = model_from_yaml(yaml_no_params)
-no_param_model.get_config(verbose=1)
-
-######################################
-# save multi-branch sequential model #
-######################################
-
-seq = Sequential()
-seq.add(Merge([model, model], mode='sum'))
-seq.get_config(verbose=1)
-merge_yaml = seq.to_yaml()
-merge_model = model_from_yaml(merge_yaml)
-
-large_model = Sequential()
-large_model.add(Merge([seq,model], mode='concat'))
-large_model.get_config(verbose=1)
-large_model.to_yaml()
-
-####################
-# save graph model #
-####################
-
-X = np.random.random((100, 32))
-X2 = np.random.random((100, 32))
-y = np.random.random((100, 4))
-y2 = np.random.random((100,))
-
-(X_train, y_train), (X_test, y_test) = get_test_data(nb_train=1000, nb_test=200, input_shape=(32,),
-                                                     classification=False, output_shape=(4,))
-
-graph = Graph()
-
-graph.add_input(name='input1', ndim=2)
-
-graph.add_node(Dense(32, 16), name='dense1', input='input1')
-graph.add_node(Dense(32, 4), name='dense2', input='input1')
-graph.add_node(Dense(16, 4), name='dense3', input='dense1')
-
-graph.add_output(name='output1', inputs=['dense2', 'dense3'], merge_mode='sum')
-graph.compile('rmsprop', {'output1': 'mse'})
-
-graph.get_config(verbose=1)
-
-history = graph.fit({'input1': X_train, 'output1': y_train}, nb_epoch=10)
-original_pred = graph.predict({'input1': X_test})
-
-graph_yaml = graph.to_yaml()
-graph.save_weights('temp.h5', overwrite=True)
-
-reloaded_graph = model_from_yaml(graph_yaml)
-reloaded_graph.load_weights('temp.h5')
-reloaded_graph.get_config(verbose=1)
-
-reloaded_graph.compile('rmsprop', {'output1': 'mse'})
-new_pred = reloaded_graph.predict({'input1': X_test})
-
-assert(np.sum(new_pred['output1'] - original_pred['output1']) == 0)

tests/test_embeddings.py

@@ -1,10 +1,10 @@
 import unittest
 import numpy as np
 from keras.models import Sequential
-from keras.layers.core import Merge, Dense, Activation, Flatten
+from keras.layers.core import Dense, Activation, Flatten
 from keras.layers.embeddings import Embedding
-from theano import function
 from keras.constraints import unitnorm
+from keras import backend as K
 
 
 class TestEmbedding(unittest.TestCase):
@@ -14,13 +14,16 @@ class TestEmbedding(unittest.TestCase):
 
     def test_unitnorm_constraint(self):
         lookup = Sequential()
-        lookup.add(Embedding(3, 2, weights=[self.W1], W_constraint=unitnorm()))
+        lookup.add(Embedding(3, 2, weights=[self.W1],
+                             W_constraint=unitnorm(),
+                             input_length=1))
         lookup.add(Flatten())
-        lookup.add(Dense(2, 1))
+        lookup.add(Dense(1))
         lookup.add(Activation('sigmoid'))
-        lookup.compile(loss='binary_crossentropy', optimizer='sgd', class_mode='binary')
+        lookup.compile(loss='binary_crossentropy', optimizer='sgd',
+                       class_mode='binary')
         lookup.train_on_batch(self.X1, np.array([[1], [0]], dtype='int32'))
-        norm = np.linalg.norm(lookup.params[0].get_value(), axis=1)
+        norm = np.linalg.norm(K.get_value(lookup.params[0]), axis=1)
         self.assertTrue(np.allclose(norm, np.ones_like(norm).astype('float32')))
 
 if __name__ == '__main__':