Commit 2dc3bccd authored by miaecle's avatar miaecle
Browse files

tensorgraph refinement

parent c7d7196c

deepchem/models/tensorgraph/graph_layers.py

+3 −3
Original line number Diff line number Diff line
@@ -336,7 +336,7 @@ class DTNNEmbedding(Layer): 


  def __init__(self,

               n_embedding=30,

               periodic_table_length=83,

               periodic_table_length=30,

               init='glorot_uniform',

               **kwargs):

    """
@@ -468,8 +468,8 @@ class DTNNGather(Layer): 


  def __init__(self,

               n_embedding=30,

               n_outputs=100,

               layer_sizes=[100],

               n_outputs=1,

               layer_sizes=[15],

               init='glorot_uniform',

               activation='tanh',

               **kwargs):

deepchem/models/tensorgraph/layers.py

+24 −23
Original line number Diff line number Diff line
@@ -489,6 +489,7 @@ class L2Loss(Layer): 
    return out_tensor







class SoftMax(Layer):



  def __init__(self, **kwargs):

deepchem/models/tensorgraph/models/graph_models.py

+8 −157
Original line number Diff line number Diff line
@@ -104,7 +104,7 @@ class WeaveTensorGraph(TensorGraph): 
        cost = L2Loss(in_layers=[label, regression])

        costs.append(cost)



    all_cost = Concat(in_layers=costs, axis=0)

    all_cost = Concat(in_layers=costs, axis=1)

    self.weights = Weights(shape=(None, self.n_tasks))

    loss = WeightedError(in_layers=[all_cost, self.weights])

    self.set_loss(loss)
@@ -118,6 +118,7 @@ class WeaveTensorGraph(TensorGraph): 
        similar to deepchem.models.tf_new_models.graph_topology.AlternateWeaveTopology.batch_to_feed_dict

        """

    for epoch in range(epochs):

      print('Starting epoch %i' % epoch)

      for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(

          batch_size=self.batch_size,

          deterministic=True,
@@ -166,42 +167,6 @@ 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)



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

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

    return self.predict_proba_on_generator(generator, transformers)



  def predict_on_generator(self, generator, transformers=[]):

    retval = self.predict_proba_on_generator(generator, transformers)

    if self.mode == 'classification':

      retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1)

    return retval



  def predict_proba_on_generator(self, generator, transformers=[]):

    if not self.built:

      self.build()

    with self._get_tf("Graph").as_default():

      with tf.Session() as sess:

        saver = tf.train.Saver()

        saver.restore(sess, self.last_checkpoint)

        out_tensors = [x.out_tensor for x in self.outputs]

        results = []

        for feed_dict in generator:

          feed_dict = {

              self.layers[k.name].out_tensor: v

              for k, v in six.iteritems(feed_dict)

          }

          result = np.array(sess.run(out_tensors, feed_dict=feed_dict))

          if len(result.shape) == 3:

            result = np.transpose(result, axes=[1, 0, 2])

          if len(transformers) > 0:

            result = undo_transforms(result, transformers)

          results.append(result)

        return np.concatenate(results, axis=0)





class DTNNTensorGraph(TensorGraph):
@@ -279,20 +244,21 @@ class DTNNTensorGraph(TensorGraph): 


    costs = []

    self.labels_fd = []

    self.weights_fd = []

    for task in range(self.n_tasks):

      regression = DTNNExtract(task, in_layers=[dtnn_gather])

      self.add_output(regression)



      label = Label(shape=(None, 1))

      self.labels_fd.append(label)

      cost = L2Loss(in_layers=[label, regression])

      costs.append(cost)



    all_cost = Concat(in_layers=costs)

    all_cost = Concat(in_layers=costs, axis=1)

    self.weights = Weights(shape=(None, self.n_tasks))

    loss = WeightedError(in_layers=[all_cost, self.weights])

    self.set_loss(loss)





  def default_generator(self,

                        dataset,

                        epochs=1,
@@ -302,6 +268,7 @@ class DTNNTensorGraph(TensorGraph): 
        similar to deepchem.models.tf_new_models.graph_topology.DTNNGraphTopology.batch_to_feed_dict

        """

    for epoch in range(epochs):

      print('Starting epoch %i' % epoch)

      for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(

          batch_size=self.batch_size,

          deterministic=True,
@@ -313,7 +280,6 @@ class DTNNTensorGraph(TensorGraph): 
            feed_dict[label] = y_b[:, index:index + 1]

        if w_b is not None and not predict:

          feed_dict[self.weights] = w_b



        distance = []

        atom_membership = []

        distance_membership_i = []
@@ -349,42 +315,6 @@ class DTNNTensorGraph(TensorGraph): 


        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)



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

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

    return self.predict_proba_on_generator(generator, transformers)



  def predict_on_generator(self, generator, transformers=[]):

    retval = self.predict_proba_on_generator(generator, transformers)

    if self.mode == 'classification':

      retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1)

    return retval



  def predict_proba_on_generator(self, generator, transformers=[]):

    if not self.built:

      self.build()

    with self._get_tf("Graph").as_default():

      with tf.Session() as sess:

        saver = tf.train.Saver()

        saver.restore(sess, self.last_checkpoint)

        out_tensors = [x.out_tensor for x in self.outputs]

        results = []

        for feed_dict in generator:

          feed_dict = {

              self.layers[k.name].out_tensor: v

              for k, v in six.iteritems(feed_dict)

          }

          result = np.array(sess.run(out_tensors, feed_dict=feed_dict))

          if len(result.shape) == 3:

            result = np.transpose(result, axes=[1, 0, 2])

          if len(transformers) > 0:

            result = undo_transforms(result, transformers)

          results.append(result)

        return np.concatenate(results, axis=0)





class DAGTensorGraph(TensorGraph):
@@ -483,6 +413,7 @@ class DAGTensorGraph(TensorGraph): 
        similar to deepchem.models.tf_new_models.graph_topology.DAGGraphTopology.batch_to_feed_dict

        """

    for epoch in range(epochs):

      print('Starting epoch %i' % epoch)

      for (X_b, y_b, w_b, ids_b) in dataset.iterbatches(

          batch_size=self.batch_size,

          deterministic=True,
@@ -529,42 +460,6 @@ class DAGTensorGraph(TensorGraph): 
        feed_dict[self.n_atoms] = n_atoms

        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)



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

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

    return self.predict_proba_on_generator(generator, transformers)



  def predict_on_generator(self, generator, transformers=[]):

    retval = self.predict_proba_on_generator(generator, transformers)

    if self.mode == 'classification':

      retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1)

    return retval



  def predict_proba_on_generator(self, generator, transformers=[]):

    if not self.built:

      self.build()

    with self._get_tf("Graph").as_default():

      with tf.Session() as sess:

        saver = tf.train.Saver()

        saver.restore(sess, self.last_checkpoint)

        out_tensors = [x.out_tensor for x in self.outputs]

        results = []

        for feed_dict in generator:

          feed_dict = {

              self.layers[k.name].out_tensor: v

              for k, v in six.iteritems(feed_dict)

          }

          result = np.array(sess.run(out_tensors, feed_dict=feed_dict))

          if len(result.shape) == 3:

            result = np.transpose(result, axes=[1, 0, 2])

          if len(transformers) > 0:

            result = undo_transforms(result, transformers)

          results.append(result)

        return np.concatenate(results, axis=0)





class GraphConvTensorGraph(TensorGraph):
@@ -650,6 +545,7 @@ class GraphConvTensorGraph(TensorGraph): 
                        predict=False,

                        pad_batches=True):

    for epoch in range(epochs):

      print('Starting epoch %i' % epoch)

      for ind, (X_b, y_b, w_b, ids_b) in enumerate(

          dataset.iterbatches(

              self.batch_size, pad_batches=True, deterministic=True)):
@@ -667,48 +563,3 @@ class GraphConvTensorGraph(TensorGraph): 
        for i in range(1, len(multiConvMol.get_deg_adjacency_lists())):

          d[self.deg_adjs[i - 1]] = multiConvMol.get_deg_adjacency_lists()[i]

        yield d
 


  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)



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

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

    return self.predict_proba_on_generator(generator, transformers)



  def predict_on_generator(self, generator, transformers=[]):

    retval = self.predict_proba_on_generator(generator, transformers)

    if self.mode == 'classification':

      retval = np.expand_dims(from_one_hot(retval, axis=2), axis=1)

    return retval



  def predict_proba_on_generator(self, generator, transformers=[]):

    if not self.built:

      self.build()

    with self._get_tf("Graph").as_default():

      with tf.Session() as sess:

        saver = tf.train.Saver()

        saver.restore(sess, self.last_checkpoint)

        out_tensors = [x.out_tensor for x in self.outputs]

        results = []

        for feed_dict in generator:

          feed_dict = {

              self.layers[k.name].out_tensor: v

              for k, v in six.iteritems(feed_dict)

          }

          result = np.array(sess.run(out_tensors, feed_dict=feed_dict))

          if len(result.shape) == 3:

            result = np.transpose(result, axes=[1, 0, 2])

          if len(transformers) > 0:

            result = undo_transforms(result, transformers)

          results.append(result)

        return np.concatenate(results, axis=0)



  def evaluate(self, dataset, metrics, transformers=[], per_task_metrics=False):

    if not self.built:

      self.build()

    return self.evaluate_generator(

        self.default_generator(dataset),

        metrics,

        labels=self.my_labels,

        weights=[self.my_task_weights])

deepchem/models/tensorgraph/symmetry_functions.py

+8 −6
Original line number Diff line number Diff line
@@ -98,11 +98,12 @@ class RadialSymmetry(Layer): 
    self.atomic_number_differentiated = atomic_number_differentiated

    self.atom_number_cases = atom_numbers

    if Rs_init is None:

      self.Rs_init = np.array([0.5, 1.17, 1.83, 2.5, 3.17, 3.83, 4.5, 5.17])

      self.Rs_init = np.array([0.5, 1.17, 1.83, 2.5, 3.17, 3.83, 4.5])

      self.Rs_init = self.Rs_init / 0.52917721092

    else:

      self.Rs_init = np.array(Rs_init)

    if ita_init is None:

      self.ita_init = np.array([4.])

      self.ita_init = np.array([1.5])

    else:

      self.ita_init = np.array(ita_init)
@@ -254,17 +255,18 @@ class AngularSymmetryMod(Layer): 
      self.lambd_init = np.array(lambd_init)



    if ita_init is None:

      self.ita_init = np.array([4.])

      self.ita_init = np.array([1.5])

    else:

      self.ita_init = np.array(ita_init)



    if zeta_init is None:

      self.zeta_init = np.array([2., 4., 8.])

      self.zeta_init = np.array([4.])

    else:

      self.zeta_init = np.array(zeta_init)



    if Rs_init is None:

      self.Rs_init = np.array([0.5, 1.17, 1.83, 2.5, 3.17, 3.83, 4.5, 5.17])

      self.Rs_init = np.array([0.5, 1.17, 1.83, 2.5, 3.17])

      self.Rs_init = self.Rs_init / 0.52917721092

    else:

      self.Rs_init = np.array(Rs_init)

deepchem/models/tensorgraph/tensor_graph.py

+2 −0
Original line number Diff line number Diff line
@@ -169,6 +169,7 @@ class TensorGraph(Model): 
            break

          if self.global_step % checkpoint_interval == checkpoint_interval - 1:

            saver.save(sess, self.save_file, global_step=self.global_step)

            self.last_checkpoint = saver.last_checkpoints[-1]

            avg_loss = float(avg_loss) / n_batches

            print('Ending global_step %d: Average loss %g' % (self.global_step,

                                                              avg_loss))
@@ -524,6 +525,7 @@ class TensorGraph(Model): 
    if self.last_checkpoint is None:

      sess.run(tf.global_variables_initializer())

      saver.save(sess, self.save_file, global_step=self.global_step)

      self.last_checkpoint = saver.last_checkpoints[-1]

    else:

      saver.restore(sess, self.last_checkpoint)

CRA Git | Maintained and supported by SUSTech CRA and CCSE