Commit 1a3c14b0 authored by miaecle's avatar miaecle
Browse files

fix bug and yapf

parent 1d002eef

README.md

+15 −8
Original line number Diff line number Diff line
@@ -496,26 +496,29 @@ Scaffold splitting 
|                |Graphconv regression|Scaffold    |0.695         |0.391         |

|                |Weave regression    |Scaffold    |0.401         |0.373         |

|qm7             |NN regression       |Index       |0.997         |0.992         |

|                |DTNN                |Index       |0.998         |0.996         |

|                |DTNN                |Index       |0.997         |0.995         |

|                |NN regression       |Random      |0.998         |0.997         |

|                |DTNN                |Random      |0.998         |0.998         |

|                |DTNN                |Random      |0.999         |0.998         |

|                |NN regression       |Stratified  |0.998         |0.997         | 

|                |DTNN                |Stratified  |0.998         |0.998         | 

|qm7b            |MT-NN regression    |Index       |0.903         |0.789         |

|                |DTNN                |Index       |0.872         |0.821         |

|                |DTNN                |Index       |0.919         |0.863         |

|                |MT-NN regression    |Random      |0.893         |0.839         |

|                |DTNN                |Random      |0.865         |0.849         |

|                |DTNN                |Random      |0.924         |0.898         |

|                |MT-NN regression    |Stratified  |0.891         |0.859         | 

|                |DTNN                |Stratified  |0.853         |0.839         | 

|                |DTNN                |Stratified  |0.913         |0.894         | 

|qm8             |MT-NN regression    |Index       |0.783         |0.656         |

|                |DTNN                |Index       |0.737         |0.639         |

|                |DTNN                |Index       |0.857         |0.691         |

|                |MT-NN regression    |Random      |0.747         |0.660         |

|                |DTNN                |Random      |0.731         |0.711         |

|                |DTNN                |Random      |0.842         |0.756         |

|                |MT-NN regression    |Stratified  |0.756         |0.681         |

|                |DTNN                |Stratified  |0.714         |0.683         | 

|                |DTNN                |Stratified  |0.844         |0.758         | 

|qm9             |MT-NN regression    |Index       |0.733         |0.766         |

|                |DTNN                |Index       |0.918         |0.831         | 

|                |MT-NN regression    |Random      |0.852         |0.833         |

|                |DTNN                |Random      |0.942         |0.948         | 

|                |MT-NN regression    |Stratified  |0.764         |0.792         | 

|                |DTNN                |Stratified  |0.941         |0.867         | 

|sampl           |Random forest       |Index       |0.968         |0.736         |

|                |XGBoost             |Index       |0.884         |0.784         |

|                |NN regression       |Index       |0.917         |0.764         |
@@ -675,9 +678,13 @@ Time needed for benchmark test(~20h in total) 
|                |Graphconv regression|20              |100            |

|                |Weave regression    |20              |120            |

|qm7             |MT-NN regression    |10              |400            |

|                |DTNN                |10              |600            |

|qm7b            |MT-NN regression    |10              |600            |

|                |DTNN                |10              |600            |

|qm8             |MT-NN regression    |60              |1000           |

|                |DTNN                |10              |2000           |

|qm9             |MT-NN regression    |220             |10000          |

|                |DTNN                |10              |14000          |

|sampl           |NN regression       |10              |30             |

|                |XGBoost             |10              |20             |

|                |Random forest       |10              |20             |

deepchem/models/tensorgraph/graph_layers.py

+27 −17
Original line number Diff line number Diff line
@@ -18,7 +18,9 @@ from deepchem.nn import model_ops 


from deepchem.models.tensorgraph.layers import Layer





class Combine_AP(Layer):



  def __init__(self, **kwargs):

    super(Combine_AP, self).__init__(**kwargs)
@@ -27,13 +29,16 @@ class Combine_AP(Layer): 
    P = self.in_layers[1].out_tensor

    self.out_tensor = [A, P]





class Separate_AP(Layer):



  def __init__(self, **kwargs):

    super(Separate_AP, self).__init__(**kwargs)



  def _create_tensor(self):

    self.out_tensor = self.in_layers[0].out_tensor[0]





class WeaveLayer(Layer):

  """ TensorGraph style implementation

  The same as deepchem.nn.WeaveLayer
@@ -173,6 +178,7 @@ class WeaveLayer(Layer): 
      P = pair_features

    self.out_tensor = [A, P]





class WeaveGather(Layer):

  """ TensorGraph style implementation

  The same as deepchem.nn.WeaveGather
@@ -233,7 +239,6 @@ class WeaveGather(Layer): 
      output_molecules = self.activation(output_molecules)

    self.out_tensor = output_molecules





  def gaussian_histogram(self, x):

    gaussian_memberships = [(-1.645, 0.283), (-1.080, 0.170), (-0.739, 0.134),

                            (-0.468, 0.118), (-0.228, 0.114), (0., 0.114),
@@ -352,6 +357,7 @@ class DTNNGather(Layer): 
  """ TensorGraph style implementation

  The same as deepchem.nn.DTNNGather

  """



  def __init__(self,

               n_embedding=30,

               n_outputs=100,
@@ -398,10 +404,12 @@ class DTNNGather(Layer): 
    output = tf.segment_sum(output, atom_membership)

    self.out_tensor = output





class DAGLayer(Layer):

  """ TensorGraph style implementation

  The same as deepchem.nn.DAGLayer

  """



  def __init__(self,

               n_graph_feat=30,

               n_atom_feat=75,
@@ -481,27 +489,28 @@ class DAGLayer(Layer): 


    n_atoms = self.in_layers[4].out_tensor

    # initialize graph features for each graph

    graph_features_initial = tf.zeros((self.max_atoms*self.batch_size, self.max_atoms+1, self.n_graph_feat))

    graph_features_initial = tf.zeros((self.max_atoms * self.batch_size,

                                       self.max_atoms + 1, self.n_graph_feat))

    # initialize graph features for each graph

    # another row of zeros is generated for padded dummy atoms

    graph_features = tf.Variable(

        graph_features_initial,

        trainable=False)

    graph_features = tf.Variable(graph_features_initial, trainable=False)



    for count in range(self.max_atoms):

      # `count`-th step

      # extracting atom features of target atoms: (batch_size*max_atoms) * n_atom_features

      mask = calculation_masks[:, count]

      current_round = tf.boolean_mask(calculation_orders[:, count], mask)

      batch_atom_features = tf.gather(atom_features,

                                      current_round)

      batch_atom_features = tf.gather(atom_features, current_round)



      # generating index for graph features used in the inputs

      index = tf.stack(

          [

              tf.reshape(

                  tf.stack([tf.boolean_mask(tf.range(n_atoms), mask)] * (self.max_atoms - 1), axis=1),

                  [-1]), tf.reshape(tf.boolean_mask(parents[:, count, 1:], mask), [-1])

                  tf.stack(

                      [tf.boolean_mask(tf.range(n_atoms), mask)] *

                      (self.max_atoms - 1),

                      axis=1), [-1]),

              tf.reshape(tf.boolean_mask(parents[:, count, 1:], mask), [-1])

          ],

          axis=1)

      # extracting graph features for parents of the target atoms, then flatten
@@ -539,6 +548,7 @@ class DAGGather(Layer): 
  """ TensorGraph style implementation

  The same as deepchem.nn.DAGGather

  """



  def __init__(self,

               n_graph_feat=30,

               n_outputs=30,

deepchem/models/tensorgraph/models/graph_models.py

+90 −65
Original line number Diff line number Diff line
@@ -11,6 +11,7 @@ from deepchem.models.tensorgraph.graph_layers import WeaveLayer, WeaveGather, \ 
from deepchem.metrics import to_one_hot, from_one_hot

from deepchem.trans import undo_transforms





class WeaveTensorGraph(TensorGraph):



  def __init__(self,
@@ -49,7 +50,8 @@ class WeaveTensorGraph(TensorGraph): 
        update_pair=False,

        in_layers=[weave_layer1, self.pair_split, self.atom_to_pair])

    separated = Separate_AP(in_layers=[weave_layer2])

    dense1 = Dense(out_channels=self.n_graph_feat, 

    dense1 = Dense(

        out_channels=self.n_graph_feat,

        activation_fn=tf.nn.relu,

        in_layers=[separated])

    batch_norm1 = BatchNormLayer(in_layers=[dense1])
@@ -63,7 +65,8 @@ class WeaveTensorGraph(TensorGraph): 
    self.labels_fd = []

    for task in range(self.n_tasks):

      if self.mode == "classification":

        classification = Dense(out_channels=2, activation_fn=None, in_layers=[weave_gather])

        classification = Dense(

            out_channels=2, activation_fn=None, in_layers=[weave_gather])

        softmax = SoftMax(in_layers=[classification])

        self.add_output(softmax)
@@ -72,7 +75,8 @@ class WeaveTensorGraph(TensorGraph): 
        cost = SoftMaxCrossEntropy(in_layers=[label, classification])

        costs.append(cost)

      if self.mode == "regression":

        regression = Dense(out_channels=1, activation_fn=None, in_layers=[weave_gather])

        regression = Dense(

            out_channels=1, activation_fn=None, in_layers=[weave_gather])

        self.add_output(regression)



        label = Label(shape=(None, 1))
@@ -119,7 +123,8 @@ class WeaveTensorGraph(TensorGraph): 
          # index of pair features

          C0, C1 = np.meshgrid(np.arange(n_atoms), np.arange(n_atoms))

          atom_to_pair.append(

              np.transpose(np.array([C1.flatten() + start, C0.flatten() + start])))

              np.transpose(

                  np.array([C1.flatten() + start, C0.flatten() + start])))

          # number of pairs for each atom

          pair_split.extend(C1.flatten() + start)

          start = start + n_atoms
@@ -138,7 +143,6 @@ class WeaveTensorGraph(TensorGraph): 
        feed_dict[self.atom_to_pair] = np.concatenate(atom_to_pair, axis=0)

        yield feed_dict





  def predict(self, dataset, transformers=[], batch_size=None):

    generator = self.default_generator(dataset, predict=True, pad_batches=False)

    return self.predict_on_generator(generator, transformers)
@@ -175,6 +179,7 @@ class WeaveTensorGraph(TensorGraph): 
          results.append(result)

        return np.concatenate(results, axis=0)





class DTNNTensorGraph(TensorGraph):



  def __init__(self,
@@ -206,15 +211,22 @@ class DTNNTensorGraph(TensorGraph): 
    self.distance_membership_i = Feature(shape=(None,), dtype=tf.int32)

    self.distance_membership_j = Feature(shape=(None,), dtype=tf.int32)



    dtnn_embedding = DTNNEmbedding(n_embedding=self.n_embedding, in_layers=[self.atom_number])

    dtnn_embedding = DTNNEmbedding(

        n_embedding=self.n_embedding, in_layers=[self.atom_number])

    dtnn_layer1 = DTNNStep(

        n_embedding=self.n_embedding,

        n_distance=self.n_distance,

      in_layers=[dtnn_embedding, self.distance, self.distance_membership_i, self.distance_membership_j])

        in_layers=[

            dtnn_embedding, self.distance, self.distance_membership_i,

            self.distance_membership_j

        ])

    dtnn_layer2 = DTNNStep(

        n_embedding=self.n_embedding,

        n_distance=self.n_distance,

      in_layers=[dtnn_layer1, self.distance, self.distance_membership_i, self.distance_membership_j])

        in_layers=[

            dtnn_layer1, self.distance, self.distance_membership_i,

            self.distance_membership_j

        ])

    dtnn_gather = DTNNGather(

        n_embedding=self.n_embedding,

        n_outputs=self.n_hidden,
@@ -223,7 +235,8 @@ class DTNNTensorGraph(TensorGraph): 
    costs = []

    self.labels_fd = []

    for task in range(self.n_tasks):

      regression = Dense(out_channels=1, activation_fn=None, in_layers=[dtnn_gather])

      regression = Dense(

          out_channels=1, activation_fn=None, in_layers=[dtnn_gather])

      self.add_output(regression)



      label = Label(shape=(None, 1))
@@ -281,13 +294,14 @@ class DTNNTensorGraph(TensorGraph): 
        distance = np.concatenate(distance, 0)

        feed_dict[self.distance] = np.exp(-np.square(distance - self.steps) /

                                          (2 * self.step_size**2))

        feed_dict[self.distance_membership_i] = np.concatenate(distance_membership_i)

        feed_dict[self.distance_membership_j] = np.concatenate(distance_membership_j)

        feed_dict[self.distance_membership_i] = np.concatenate(

            distance_membership_i)

        feed_dict[self.distance_membership_j] = np.concatenate(

            distance_membership_j)

        feed_dict[self.atom_membership] = np.concatenate(atom_membership)



        yield feed_dict





  def predict(self, dataset, transformers=[], batch_size=None):

    generator = self.default_generator(dataset, predict=True, pad_batches=False)

    return self.predict_on_generator(generator, transformers)
@@ -324,6 +338,7 @@ class DTNNTensorGraph(TensorGraph): 
          results.append(result)

        return np.concatenate(results, axis=0)





class DAGTensorGraph(TensorGraph):



  def __init__(self,
@@ -343,9 +358,12 @@ class DAGTensorGraph(TensorGraph): 


  def build_graph(self):

    self.atom_features = Feature(shape=(None, self.n_atom_feat))

    self.parents = Feature(shape=(None, self.max_atoms, self.max_atoms), dtype=tf.int32)

    self.calculation_orders = Feature(shape=(None, self.max_atoms), dtype=tf.int32)

    self.calculation_masks = Feature(shape=(None, self.max_atoms), dtype=tf.bool)

    self.parents = Feature(

        shape=(None, self.max_atoms, self.max_atoms), dtype=tf.int32)

    self.calculation_orders = Feature(

        shape=(None, self.max_atoms), dtype=tf.int32)

    self.calculation_masks = Feature(

        shape=(None, self.max_atoms), dtype=tf.bool)

    self.membership = Feature(shape=(None,), dtype=tf.int32)

    self.n_atoms = Feature(shape=(), dtype=tf.int32)

    dag_layer1 = DAGLayer(
@@ -353,7 +371,10 @@ class DAGTensorGraph(TensorGraph): 
        n_atom_feat=self.n_atom_feat,

        max_atoms=self.max_atoms,

        batch_size=self.batch_size,

        in_layers=[self.atom_features, self.parents, self.calculation_orders, self.calculation_masks, self.n_atoms])

        in_layers=[

            self.atom_features, self.parents, self.calculation_orders,

            self.calculation_masks, self.n_atoms

        ])

    dag_gather = DAGGather(

        n_graph_feat=self.n_graph_feat,

        n_outputs=self.n_outputs,
@@ -364,7 +385,8 @@ class DAGTensorGraph(TensorGraph): 
    self.labels_fd = []

    for task in range(self.n_tasks):

      if self.mode == "classification":

        classification = Dense(out_channels=2, activation_fn=None, in_layers=[dag_gather])

        classification = Dense(

            out_channels=2, activation_fn=None, in_layers=[dag_gather])

        softmax = SoftMax(in_layers=[classification])

        self.add_output(softmax)
@@ -373,7 +395,8 @@ class DAGTensorGraph(TensorGraph): 
        cost = SoftMaxCrossEntropy(in_layers=[label, classification])

        costs.append(cost)

      if self.mode == "regression":

        regression = Dense(out_channels=1, activation_fn=None, in_layers=[dag_gather])

        regression = Dense(

            out_channels=1, activation_fn=None, in_layers=[dag_gather])

        self.add_output(regression)



        label = Label(shape=(None, 1))
@@ -430,8 +453,10 @@ class DAGTensorGraph(TensorGraph): 


        feed_dict[self.atom_features] = np.concatenate(atoms_all, axis=0)

        feed_dict[self.parents] = np.stack(parents_all, axis=0)

        feed_dict[self.calculation_orders] = np.concatenate(calculation_orders, axis=0)

        feed_dict[self.calculation_masks] = np.concatenate(calculation_masks, axis=0)

        feed_dict[self.calculation_orders] = np.concatenate(

            calculation_orders, axis=0)

        feed_dict[self.calculation_masks] = np.concatenate(

            calculation_masks, axis=0)

        feed_dict[self.membership] = np.array(membership)

        feed_dict[self.n_atoms] = n_atoms

        yield feed_dict

deepchem/models/tests/test_overfit.py

+3 −5
Original line number Diff line number Diff line
@@ -677,10 +677,9 @@ class TestOverfit(test_util.TensorFlowTestCase): 
    regression_metric = dc.metrics.Metric(

        dc.metrics.pearson_r2_score, task_averager=np.mean)

    n_tasks = y.shape[1]

    max_n_atoms = list(dataset.get_data_shape())[0]

    batch_size = 10



    graph_model = dc.nn.SequentialDTNNGraph(max_n_atoms=max_n_atoms)

    graph_model = dc.nn.SequentialDTNNGraph()

    graph_model.add(dc.nn.DTNNEmbedding(n_embedding=20))

    graph_model.add(dc.nn.DTNNStep(n_embedding=20))

    graph_model.add(dc.nn.DTNNStep(n_embedding=20))
@@ -728,9 +727,8 @@ class TestOverfit(test_util.TensorFlowTestCase): 
    transformer = dc.trans.DAGTransformer(max_atoms=50)

    dataset = transformer.transform(dataset)



    graph = dc.nn.SequentialDAGGraph(

        n_feat, batch_size=batch_size, max_atoms=50)

    graph.add(dc.nn.DAGLayer(30, n_feat, max_atoms=50))

    graph = dc.nn.SequentialDAGGraph(n_atom_feat=n_feat, max_atoms=50)

    graph.add(dc.nn.DAGLayer(30, n_feat, max_atoms=50, batch_size=batch_size))

    graph.add(dc.nn.DAGGather(max_atoms=50))



    model = dc.models.MultitaskGraphRegressor(

deepchem/models/tf_new_models/graph_models.py

+2 −1
Original line number Diff line number Diff line
@@ -149,7 +149,8 @@ class SequentialDAGGraph(SequentialGraph): 
        self.output = layer([self.output] +

                            self.graph_topology.get_topology_placeholders())

      elif type(layer).__name__ in ['DAGGather']:

        self.output = layer([self.output, self.graph_topology.membership_placeholder])

        self.output = layer(

            [self.output, self.graph_topology.membership_placeholder])

      else:

        self.output = layer(self.output)

      self.layers.append(layer)

CRA Git | Maintained and supported by SUSTech CRA and CCSE