Commented out obsolete code

import tensorflow as tf

from deepchem.models.tensorgraph.tensor_graph import MultiTaskTensorGraph

from deepchem.models.tensorgraph.layers import Input, Dense, Concat, SoftMax, SoftMaxCrossEntropy, Layer

class WeightedError(Layer):

  def __call__(self, *parents):

    entropy, weights = parents[0], parents[1]

    self.out_tensor = tf.reduce_sum(entropy.out_tensor * weights.out_tensor)

    return self.out_tensor

def tensorGraphMultitaskClassifier(n_tasks,

                                   n_features,

                                   layer_sizes=[500],

                                   bypass_layer_sizes=[100],

                                   model_dir=None):

  """

  TODO(LESWING) Add Dropout and regularization

  Parameters

  ----------

  n_tasks

  n_features

  layer_sizes

  bypass_layer_sizes

  model_dir

  Returns

  -------

  """

  g = MultiTaskTensorGraph(model_dir=model_dir)

  in_layer = Input(shape=(None, n_features), name="FEATURE")

  g.add_layer(in_layer)

  g.add_feature(in_layer)

  # Shared Dense Layers

  prev_layer = in_layer

  dense_layers = []

  for i in range(len(layer_sizes)):

    dense = Dense(

        out_channels=layer_sizes[i],

        name="SDENSE%s" % i,

        activation_fn=tf.nn.relu)

    g.add_layer(dense, parents=[prev_layer])

    dense_layers.append(dense)

    prev_layer = dense

  # Individual Bypass Layers

  costs = []

  for task in range(n_tasks):

    prev_layer = in_layer

    for i in range(len(bypass_layer_sizes)):

      dense = Dense(

          out_channels=bypass_layer_sizes[i], name="BDENSE%s_%s" % (i, task))

      g.add_layer(dense, parents=[prev_layer])

      prev_layer = dense

    joined_layer = Concat(name="JOIN%s" % task)

    g.add_layer(joined_layer, parents=[dense_layers[-1], prev_layer])

    classification = Dense(out_channels=2, name="GUESS%s" % task)

    g.add_layer(classification, parents=[joined_layer])

    softmax = SoftMax(name="SOFTMAX%s" % task)

    g.add_layer(softmax, parents=[classification])

    g.add_output(softmax)

    label = Input(shape=(None, 2), name="LABEL%s" % task)

    g.add_layer(label)

    g.add_label(label)

    cost = SoftMaxCrossEntropy(name="COST%s" % task)

    g.add_layer(cost, parents=[label, classification])

    costs.append(cost)

  entropy = Concat(name="ENT")

  g.add_layer(entropy, parents=costs)

  task_weights = Input(shape=(None, n_tasks), name="W")

  g.add_layer(task_weights)

  g.set_task_weights(task_weights)

  loss = WeightedError(name="ERROR")

  g.add_layer(loss, parents=[entropy, task_weights])

  g.set_loss(loss)

  return g

#import tensorflow as tf

#from deepchem.models.tensorgraph.tensor_graph import MultiTaskTensorGraph

#from deepchem.models.tensorgraph.layers import Input, Dense, Concat, SoftMax, SoftMaxCrossEntropy, Layer

#

#

#class WeightedError(Layer):

#

#  def __call__(self, *parents):

#    entropy, weights = parents[0], parents[1]

#    self.out_tensor = tf.reduce_sum(entropy.out_tensor * weights.out_tensor)

#    return self.out_tensor

#

#

#def tensorGraphMultitaskClassifier(n_tasks,

#                                   n_features,

#                                   layer_sizes=[500],

#                                   bypass_layer_sizes=[100],

#                                   model_dir=None):

#  """

#  TODO(LESWING) Add Dropout and regularization

#  

#  Parameters

#  ----------

#  n_tasks

#  n_features

#  layer_sizes

#  bypass_layer_sizes

#  model_dir

#

#  Returns

#  -------

#

#  """

#  g = MultiTaskTensorGraph(model_dir=model_dir)

#  in_layer = Input(shape=(None, n_features), name="FEATURE")

#  g.add_layer(in_layer)

#  g.add_feature(in_layer)

#

#  # Shared Dense Layers

#  prev_layer = in_layer

#  dense_layers = []

#  for i in range(len(layer_sizes)):

#    dense = Dense(

#        out_channels=layer_sizes[i],

#        name="SDENSE%s" % i,

#        activation_fn=tf.nn.relu)

#    g.add_layer(dense, parents=[prev_layer])

#    dense_layers.append(dense)

#    prev_layer = dense

#

#  # Individual Bypass Layers

#  costs = []

#  for task in range(n_tasks):

#    prev_layer = in_layer

#    for i in range(len(bypass_layer_sizes)):

#      dense = Dense(

#          out_channels=bypass_layer_sizes[i], name="BDENSE%s_%s" % (i, task))

#      g.add_layer(dense, parents=[prev_layer])

#      prev_layer = dense

#    joined_layer = Concat(name="JOIN%s" % task)

#    g.add_layer(joined_layer, parents=[dense_layers[-1], prev_layer])

#

#    classification = Dense(out_channels=2, name="GUESS%s" % task)

#    g.add_layer(classification, parents=[joined_layer])

#

#    softmax = SoftMax(name="SOFTMAX%s" % task)

#    g.add_layer(softmax, parents=[classification])

#    g.add_output(softmax)

#

#    label = Input(shape=(None, 2), name="LABEL%s" % task)

#    g.add_layer(label)

#    g.add_label(label)

#

#    cost = SoftMaxCrossEntropy(name="COST%s" % task)

#    g.add_layer(cost, parents=[label, classification])

#    costs.append(cost)

#

#  entropy = Concat(name="ENT")

#  g.add_layer(entropy, parents=costs)

#

#  task_weights = Input(shape=(None, n_tasks), name="W")

#  g.add_layer(task_weights)

#  g.set_task_weights(task_weights)

#

#  loss = WeightedError(name="ERROR")

#  g.add_layer(loss, parents=[entropy, task_weights])

#  g.set_loss(loss)

#

#  return g