Merge pull request #356 from joegomes/tf12

- pip install keras

- export KERAS_BACKEND=tensorflow

- conda install -c omnia mdtraj

- pip install tensorflow

- python setup.py install

#- if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.12.0rc1-cp27-none-linux_x86_64.whl;

#  -O else export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.12.0rc1-cp35-cp35m-linux_x86_64.whl;

#  -O fi 

#- if [[ "$TRAVIS_PYTHON_VERSION" == "2.7" ]]; then pip install --ignore-installed --upgrade $TF_BINARY_URL;

#  -O else pip3 install --ignore-installed --upgrade $TF_BINARY_URL;

#  -O fi

- conda install -c https://conda.anaconda.org/jjhelmus tensorflow=0.10.0rc0

script:

- nosetests -v deepchem --nologcapture

after_success:

      train_op = self.get_training_op(

          self.train_graph.graph, self.train_graph.loss)

      with self._get_shared_session(train=True) as sess:

        sess.run(tf.initialize_all_variables())

        sess.run(tf.global_variables_initializer())

        saver = tf.train.Saver(max_to_keep=max_checkpoints_to_keep)

        # Save an initial checkpoint.

        saver.save(sess, self._save_path, global_step=0)

      # self._save_path is "logdir/model.ckpt"

      if os.path.basename(self._save_path) in filename:

        try:

          N = int(filename.split("-")[-1])

          N = int(filename.split("-")[1].split(".")[0])

          if N > highest_num:

            highest_num = N

            last_checkpoint = filename

            last_checkpoint = "model.ckpt-"+str(N)

        except ValueError:

          pass

    return os.path.join(self.logdir, last_checkpoint)

  """

  if init is None:

    init = tf.zeros([tensor.get_shape()[-1].value])

  with tf.op_scope([tensor], name, tensor.op.name):

  with tf.name_scope(name, tensor.op.name, [tensor]):

    b = tf.Variable(init, name='b')

    return tf.nn.bias_add(tensor, b)

  if bias_init is None:

    bias_init = tf.zeros([size])

  with tf.op_scope([tensor], name, 'fully_connected'):

  with tf.name_scope(name, 'fully_connected', [tensor]):

    w = tf.Variable(weight_init, name='w', dtype=tf.float32)

    b = tf.Variable(bias_init, name='b', dtype=tf.float32)

    return tf.nn.xw_plus_b(tensor, w, b)

  logits_list = []

  with tf.name_scope('multitask_logits'):

    for task_idx in range(num_tasks):

      with tf.op_scope([features], name,

                       ('task' + str(task_idx).zfill(len(str(num_tasks))))):

      with tf.name_scope(name,

                       ('task' + str(task_idx).zfill(len(str(num_tasks)))), [features]):

        logits_list.append(

            logits(features, num_classes, weight_init=weight_init,

                   bias_init=bias_init, dropout_prob=dropout_prob))

  Returns:

    A logits tensor with shape batch_size x num_classes.

  """

  with tf.op_scope([features], name, 'logits') as name:

  with tf.name_scope(name, 'logits', [features]) as name:

    return dropout(

        fully_connected_layer(features, num_classes, weight_init=weight_init,

                              bias_init=bias_init, name=name),

  Returns:

    A tensor with softmax-normalized values on the last dimension.

  """

  with tf.op_scope([tensor], name, 'softmax_N'):

  with tf.name_scope(name, 'softmax_N', [tensor]):

    exp_tensor = tf.exp(tensor)

    reduction_indices = [tensor.get_shape().ndims - 1]

    return tf.div(exp_tensor,

      train_op = self.get_training_op(

          self.train_graph.graph, self.train_graph.loss)

      with self._get_shared_session(train=True) as sess:

        sess.run(tf.initialize_all_variables())

        sess.run(tf.global_variables_initializer())

        saver = tf.train.Saver(max_to_keep=max_checkpoints_to_keep)

        # Save an initial checkpoint.

        saver.save(sess, self._save_path, global_step=0)