Loading deepchem/models/tensorgraph/layers.py +225 −122 Original line number Diff line number Diff line Loading @@ -96,13 +96,13 @@ def convert_to_layers(in_layers): raise ValueError("convert_to_layers must be invoked on layers or tensors") return layers class Conv1DLayer(Layer): class Conv1D(Layer): def __init__(self, width, out_channels, **kwargs): self.width = width self.out_channels = out_channels self.out_tensor = None super(Conv1DLayer, self).__init__(**kwargs) super(Conv1D, self).__init__(**kwargs) def create_tensor(self, in_layers=None): if in_layers is None: Loading Loading @@ -190,6 +190,7 @@ class Dense(Layer): class Flatten(Layer): """Flatten every dimension except the first""" def __init__(self, **kwargs): super(Flatten, self).__init__(**kwargs) Loading @@ -216,21 +217,28 @@ class Reshape(Layer): self.shape = shape super(Reshape, self).__init__(**kwargs) def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor self.out_tensor = tf.reshape(parent_tensor, self.shape) return self.out_tensor class Transpose(Layer): def __init__(self, out_shape, **kwargs): def __init__(self, perm, **kwargs): super(Transpose, self).__init__(**kwargs) self.out_shape = out_shape self.perm = perm def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Only One Parent to Transpose over") self.out_tensor = tf.transpose(self.in_layers[0].out_tensor, self.out_shape) self.out_tensor = tf.transpose(in_layers[0].out_tensor, self.perm) return self.out_tensor Loading @@ -239,14 +247,18 @@ class CombineMeanStd(Layer): def __init__(self, **kwargs): super(CombineMeanStd, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: raise ValueError("Must have two self.in_layers") mean_parent, std_parent = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError("Must have two in_layers") mean_parent, std_parent = in_layers[0], in_layers[1] mean_parent_tensor, std_parent_tensor = mean_parent.out_tensor, std_parent.out_tensor sample_noise = tf.random_normal( mean_parent_tensor.get_shape(), 0, 1, dtype=tf.float32) self.out_tensor = mean_parent_tensor + (std_parent_tensor * sample_noise) return self.out_tensor class Repeat(Layer): Loading @@ -255,13 +267,17 @@ class Repeat(Layer): self.n_times = n_times super(Repeat, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor parent_tensor = in_layers[0].out_tensor t = tf.expand_dims(parent_tensor, 1) pattern = tf.stack([1, self.n_times, 1]) self.out_tensor = tf.tile(t, pattern) return self.out_tensor class GRU(Layer): Loading @@ -272,11 +288,14 @@ class GRU(Layer): self.batch_size = batch_size super(GRU, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor gru_cell = tf.nn.rnn_cell.GRUCell(self.n_hidden) parent_tensor = in_layers[0].out_tensor gru_cell = tf.contrib.rnn.GRUCell(self.n_hidden) initial_gru_state = gru_cell.zero_state(self.batch_size, tf.float32) rnn_outputs, rnn_states = tf.nn.dynamic_rnn( gru_cell, Loading @@ -285,20 +304,27 @@ class GRU(Layer): scope=self.name) projection = lambda x: tf.contrib.layers.linear(x, num_outputs=self.out_channels, activation_fn=tf.nn.sigmoid) self.out_tensor = tf.map_fn(projection, rnn_outputs) return self.out_tensor class TimeSeriesDense(Layer): def __init__(self, out_channels, **kwargs): self.out_channels = out_channels super(TimeSeriesDense, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor dense_fn = lambda x: tf.contrib.layers.fully_connected(x, num_outputs=self.out_channels, parent_tensor = in_layers[0].out_tensor dense_fn = lambda x: tf.contrib.layers.fully_connected( x, num_outputs=self.out_channels, activation_fn=tf.nn.sigmoid) self.out_tensor = tf.map_fn(dense_fn, parent_tensor) return self.out_tensor class Input(Layer): Loading @@ -309,9 +335,12 @@ class Input(Layer): super(Input, self).__init__(**kwargs) self.op_type = "cpu" def _create_tensor(self): if len(self.in_layers) > 0: queue = self.in_layers[0] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 0: queue = in_layers[0] placeholder = queue.out_tensors[self.get_pre_q_name()] self.out_tensor = tf.placeholder_with_default(placeholder, self.shape) return self.out_tensor Loading Loading @@ -344,13 +373,16 @@ class Weights(Input): super(Weights, self).__init__(**kwargs) class L2LossLayer(Layer): class L2Loss(Layer): def __init__(self, **kwargs): super(L2LossLayer, self).__init__(**kwargs) super(L2Loss, self).__init__(**kwargs) def _create_tensor(self): guess, label = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) guess, label = in_layers[0], in_layers[1] self.out_tensor = tf.reduce_mean( tf.square(guess.out_tensor - label.out_tensor)) return self.out_tensor Loading @@ -361,10 +393,13 @@ class SoftMax(Layer): def __init__(self, **kwargs): super(SoftMax, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must only Softmax single parent") parent = self.in_layers[0] parent = in_layers[0] self.out_tensor = tf.contrib.layers.softmax(parent.out_tensor) return self.out_tensor Loading @@ -375,11 +410,14 @@ class Concat(Layer): self.axis = axis super(Concat, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) == 1: self.out_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) == 1: self.out_tensor = in_layers[0].out_tensor return self.out_tensor out_tensors = [x.out_tensor for x in self.in_layers] out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.concat(out_tensors, axis=self.axis) return self.out_tensor Loading @@ -394,11 +432,13 @@ class InteratomicL2Distances(Layer): self.ndim = ndim super(InteratomicL2Distances, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError("InteratomicDistances requires coords,nbr_list") coords, nbr_list = (self.in_layers[0].out_tensor, self.in_layers[1].out_tensor) coords, nbr_list = (in_layers[0].out_tensor, in_layers[1].out_tensor) N_atoms, M_nbrs, ndim = self.N_atoms, self.M_nbrs, self.ndim # Shape (N_atoms, M_nbrs, ndim) nbr_coords = tf.gather(coords, nbr_list) Loading @@ -416,10 +456,13 @@ class SoftMaxCrossEntropy(Layer): def __init__(self, **kwargs): super(SoftMaxCrossEntropy, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError() labels, logits = self.in_layers[0].out_tensor, self.in_layers[1].out_tensor labels, logits = in_layers[0].out_tensor, in_layers[1].out_tensor self.out_tensor = tf.nn.softmax_cross_entropy_with_logits( logits=logits, labels=labels) self.out_tensor = tf.reshape(self.out_tensor, [-1, 1]) Loading @@ -432,12 +475,15 @@ class ReduceMean(Layer): self.axis = axis super(ReduceMean, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) > 1: out_tensors = [x.out_tensor for x in self.in_layers] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.stack(out_tensors) else: self.out_tensor = self.in_layers[0].out_tensor self.out_tensor = in_layers[0].out_tensor self.out_tensor = tf.reduce_mean(self.out_tensor) return self.out_tensor Loading @@ -445,10 +491,13 @@ class ReduceMean(Layer): class ToFloat(Layer): def _create_tensor(self): if len(self.in_layers) > 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: raise ValueError("Only one layer supported.") self.out_tensor = tf.to_float(self.in_layers[0].out_tensor) self.out_tensor = tf.to_float(in_layers[0].out_tensor) return self.out_tensor Loading @@ -458,12 +507,15 @@ class ReduceSum(Layer): self.axis = axis super(ReduceSum, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) > 1: out_tensors = [x.out_tensor for x in self.in_layers] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.stack(out_tensors) else: self.out_tensor = self.in_layers[0].out_tensor self.out_tensor = in_layers[0].out_tensor self.out_tensor = tf.reduce_sum(self.out_tensor, axis=self.axis) return self.out_tensor Loading @@ -475,23 +527,29 @@ class ReduceSquareDifference(Layer): self.axis = axis super(ReduceSquareDifference, self).__init__(**kwargs) def _create_tensor(self): a = self.in_layers[0].out_tensor b = self.in_layers[1].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) a = in_layers[0].out_tensor b = in_layers[1].out_tensor self.out_tensor = tf.reduce_mean( tf.squared_difference(a, b), axis=self.axis) return self.out_tensor class Conv2d(Layer): class Conv2D(Layer): def __init__(self, num_outputs, kernel_size=5, **kwargs): self.num_outputs = num_outputs self.kernel_size = kernel_size super(Conv2d, self).__init__(**kwargs) super(Conv2D, self).__init__(**kwargs) def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor out_tensor = tf.contrib.layers.conv2d( parent_tensor, num_outputs=self.num_outputs, Loading @@ -500,6 +558,7 @@ class Conv2d(Layer): activation_fn=tf.nn.relu, normalizer_fn=tf.contrib.layers.batch_norm) self.out_tensor = out_tensor return self.out_tensor class MaxPool(Layer): Loading @@ -514,8 +573,11 @@ class MaxPool(Layer): self.padding = padding super(MaxPool, self).__init__(**kwargs) def _create_tensor(self): in_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) in_tensor = in_layers[0].out_tensor self.out_tensor = tf.nn.max_pool( in_tensor, ksize=self.ksize, strides=self.strides, padding=self.padding) return self.out_tensor Loading @@ -535,12 +597,15 @@ class InputFifoQueue(Layer): super(InputFifoQueue, self).__init__(**kwargs) self.op_type = "cpu" def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if self.dtypes is None: self.dtypes = [tf.float32] * len(self.shapes) self.queue = tf.FIFOQueue( self.capacity, self.dtypes, shapes=self.shapes, names=self.names) feed_dict = {x.name: x.out_tensor for x in self.in_layers} feed_dict = {x.out_tensor.name: x.out_tensor for x in in_layers} self.out_tensor = self.queue.enqueue(feed_dict) self.close_op = self.queue.close() self.out_tensors = self.queue.dequeue() Loading @@ -557,7 +622,7 @@ class InputFifoQueue(Layer): self.queue.close() class GraphConvLayer(Layer): class GraphConv(Layer): def __init__(self, out_channel, Loading @@ -570,11 +635,14 @@ class GraphConvLayer(Layer): self.max_degree = max_deg self.num_deg = 2 * max_deg + (1 - min_deg) self.activation_fn = activation_fn super(GraphConvLayer, self).__init__(**kwargs) super(GraphConv, self).__init__(**kwargs) def _create_tensor(self): # self.in_layers = [atom_features, deg_slice, membership, deg_adj_list placeholders...] in_channels = self.in_layers[0].out_tensor.get_shape()[-1].value def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) # in_layers = [atom_features, deg_slice, membership, deg_adj_list placeholders...] in_channels = in_layers[0].out_tensor.get_shape()[-1].value # Generate the nb_affine weights and biases self.W_list = [ Loading @@ -588,11 +656,11 @@ class GraphConvLayer(Layer): ] # Extract atom_features atom_features = self.in_layers[0].out_tensor atom_features = in_layers[0].out_tensor # Extract graph topology deg_slice = self.in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in self.in_layers[3:]] deg_slice = in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in in_layers[3:]] # Perform the mol conv # atom_features = graph_conv(atom_features, deg_adj_lists, deg_slice, Loading Loading @@ -668,17 +736,21 @@ class GraphConvLayer(Layer): self.out_tensor, self.W_list, self.b_list = tensors class GraphPoolLayer(Layer): class GraphPool(Layer): def __init__(self, min_degree=0, max_degree=10, **kwargs): self.min_degree = min_degree self.max_degree = max_degree super(GraphPoolLayer, self).__init__(**kwargs) super(GraphPool, self).__init__(**kwargs) def _create_tensor(self): atom_features = self.in_layers[0].out_tensor deg_slice = self.in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in self.in_layers[3:]] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) atom_features = in_layers[0].out_tensor deg_slice = in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in in_layers[3:]] # Perform the mol gather # atom_features = graph_pool(atom_features, deg_adj_lists, deg_slice, Loading Loading @@ -721,12 +793,16 @@ class GraphGather(Layer): self.activation_fn = activation_fn super(GraphGather, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) # x = [atom_features, deg_slice, membership, deg_adj_list placeholders...] atom_features = self.in_layers[0].out_tensor atom_features = in_layers[0].out_tensor # Extract graph topology membership = self.in_layers[2].out_tensor membership = in_layers[2].out_tensor # Perform the mol gather Loading Loading @@ -757,44 +833,55 @@ class GraphGather(Layer): return mol_features class BatchNormLayer(Layer): class BatchNorm(Layer): def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor self.out_tensor = tf.layers.batch_normalization(parent_tensor) return self.out_tensor class WeightedError(Layer): def _create_tensor(self): entropy, weights = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) entropy, weights = in_layers[0], in_layers[1] self.out_tensor = tf.reduce_sum(entropy.out_tensor * weights.out_tensor) return self.out_tensor class Cutoff(Layer): """Truncates interactions that are too far away.""" def _create_tensor(self): if len(self.in_layers) != 2: raise ValueError("Cutoff must be given distances and energies.") d, x = self.in_layers[0].out_tensor, self.in_layers[1].out_tensor self.out_tensor = tf.where(d < 8, x, tf.zeros_like(x)) return self.out_tensor class VinaFreeEnergy(Layer): """Computes free-energy as defined by Autodock Vina.""" """Computes free-energy as defined by Autodock Vina. def __init__(self, stddev=.3, Nrot=1, **kwargs): TODO(rbharath): Make this layer support batching. """ def __init__(self, N_atoms, M_nbrs, ndim, nbr_cutoff, start, stop, stddev=.3, Nrot=1, **kwargs): self.stddev = stddev # Number of rotatable bonds # TODO(rbharath): Vina actually sets this per-molecule. See if makes # a difference. self.Nrot = Nrot self.N_atoms = N_atoms self.M_nbrs = M_nbrs self.ndim = ndim self.nbr_cutoff = nbr_cutoff self.start = start self.stop = stop super(VinaFreeEnergy, self).__init__(**kwargs) def cutoff(self, d, x): out_tensor = tf.where(d < 8, x, tf.zeros_like(x)) return out_tensor def nonlinearity(self, c): """Computes non-linearity used in Vina.""" w = tf.Variable(tf.random_normal((1,), stddev=self.stddev)) Loading Loading @@ -836,13 +923,13 @@ class VinaFreeEnergy(Layer): out_tensor = tf.exp(-((d - 3) / 2)**2) return out_tensor def _create_tensor(self): def create_tensor(self, in_layers=None): """ Parameters ---------- X: tf.Tensor of shape (B, N, d) X: tf.Tensor of shape (N, d) Coordinates/features. Z: tf.Tensor of shape (B, N) Z: tf.Tensor of shape (N) Atomic numbers of neighbor atoms. Returns Loading @@ -850,16 +937,21 @@ class VinaFreeEnergy(Layer): layer: tf.Tensor of shape (B) The free energy of each complex in batch """ X = self.in_layers[0].out_tensor Z = self.in_layers[2].out_tensor if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) X = in_layers[0].out_tensor Z = in_layers[1].out_tensor # TODO(rbharath): This layer shouldn't be neighbor-listing. Make # neighbors lists an argument instead of a part of this layer. nbr_list = NeighborList( self.N_atoms, self.M_nbrs, self.ndim, self.nbr_cutoff, self.start, self.stop)(coords) self.stop)(X) # Shape (N, M) dists = InteratomicL2Distances( self.N_atoms, self.M_nbrs, self.ndim)(coords, nbr_list) self.N_atoms, self.M_nbrs, self.ndim)(X, nbr_list) repulsion = self.repulsion(dists) hydrophobic = self.hydrophobic(dists) Loading @@ -872,12 +964,12 @@ class VinaFreeEnergy(Layer): repulsion, hydrophobic, hbond, gauss_1, gauss_2) # Shape (N, M) thresholded = Cutoff()(dists, interactions) thresholded = self.cutoff(dists, interactions) free_energies = self.nonlinearity(thresholded) free_energy = ReduceSum()(free_energies) self.output_tensor = free_energy self.out_tensor = free_energy return self.out_tensor Loading @@ -888,10 +980,13 @@ class WeightedLinearCombo(Layer): self.std = std super(WeightedLinearCombo, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) weights = [] out_tensor = None for in_layer in self.in_layers: for in_layer in in_layers: w = tf.Variable(tf.random_normal([ 1, ], stddev=self.std)) Loading @@ -908,6 +1003,8 @@ class NeighborList(Layer): Neighbor-lists (also called Verlet Lists) are a tool for grouping atoms which are close to each other spatially TODO(rbharath): Make this layer support batching. """ def __init__(self, N_atoms, M_nbrs, ndim, nbr_cutoff, start, stop, **kwargs): Loading Loading @@ -935,12 +1032,15 @@ class NeighborList(Layer): self.stop = stop super(NeighborList, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): """Creates tensors associated with neighbor-listing.""" if len(self.in_layers) != 1: if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Only One Parent to NeighborList over %s" % self.in_layers) parent = self.in_layers[0] in_layers) parent = in_layers[0] if len(parent.out_tensor.get_shape()) != 2: # TODO(rbharath): Support batching raise ValueError("Parent tensor must be (num_atoms, ndum)") Loading Loading @@ -1209,7 +1309,7 @@ class AtomicConvolution(Layer): self.atom_types = atom_types super(AtomicConvolution, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): """ Parameters ---------- Loading @@ -1225,10 +1325,13 @@ class AtomicConvolution(Layer): layer: tf.Tensor of shape (l, B, N) A new tensor representing the output of the atomic conv layer """ if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) X = self.in_layers[0].out_tensor Nbrs = tf.to_int32(self.in_layers[1].out_tensor) Nbrs_Z = self.in_layers[2].out_tensor X = in_layers[0].out_tensor Nbrs = tf.to_int32(in_layers[1].out_tensor) Nbrs_Z = in_layers[2].out_tensor # N: Maximum number of atoms # M: Maximum number of neighbors Loading deepchem/models/tensorgraph/models/graph_conv.py +1 −1 Original line number Diff line number Diff line import tensorflow as tf from deepchem.models.tensorgraph.tensor_graph import TensorGraph from deepchem.models.tensorgraph.layers import Input, Dense, Concat, SoftMax, SoftMaxCrossEntropy, Layer, \ GraphConvLayer, BatchNormLayer, GraphPoolLayer, GraphGather, WeightedError GraphConv, BatchNormLayer, GraphPoolLayer, GraphGather, WeightedError from deepchem.metrics import to_one_hot from deepchem.feat.mol_graphs import ConvMol import time Loading deepchem/models/tensorgraph/models/graph_models.py +2 −2 Original line number Diff line number Diff line Loading @@ -4,8 +4,8 @@ import six from deepchem.models.tensorgraph.tensor_graph import TensorGraph from deepchem.utils.evaluate import GeneratorEvaluator from deepchem.models.tensorgraph.layers import Input, BatchNormLayer, Dense, \ SoftMax, SoftMaxCrossEntropy, L2LossLayer, Concat, WeightedError, Label, Weights, Feature from deepchem.models.tensorgraph.layers import Input, BatchNorm, Dense, \ SoftMax, SoftMaxCrossEntropy, L2Loss, Concat, WeightedError, Label, Weights, Feature from deepchem.models.tensorgraph.graph_layers import WeaveLayer, WeaveGather, \ Combine_AP, Separate_AP, DTNNEmbedding, DTNNStep, DTNNGather, DAGLayer, DAGGather from deepchem.metrics import to_one_hot, from_one_hot Loading deepchem/models/tensorgraph/tensor_graph.py +1 −1 Original line number Diff line number Diff line Loading @@ -317,7 +317,7 @@ class TensorGraph(Model): for node in order: with tf.name_scope(node): node_layer = self.layers[node] node_layer._create_tensor() node_layer.create_tensor() self.built = True for layer in self.layers.values(): Loading deepchem/models/tensorgraph/tests/test_docking.py +0 −19 Original line number Diff line number Diff line Loading @@ -470,22 +470,3 @@ class TestDocking(test_util.TensorFlowTestCase): tg = dc.models.TensorGraph(learning_rate=0.1, use_queue=False) tg.set_loss(loss) tg.fit_generator(databag.iterbatches(epochs=1)) def test_interatomic_distances(self): """Test that the interatomic distance calculation works.""" N_atoms = 5 M_nbrs = 2 ndim = 3 with self.test_session() as sess: coords = np.random.rand(N_atoms, ndim) nbr_list = np.random.randint(0, N_atoms, size=(N_atoms, M_nbrs)) coords_tensor = tf.convert_to_tensor(coords) nbr_list_tensor = tf.convert_to_tensor(nbr_list) dist_tensor = InteratomicL2Distances(N_atoms, M_nbrs, ndim)( coords_tensor, nbr_list_tensor) dists = dist_tensor.eval() assert dists.shape == (N_atoms, M_nbrs) Loading
deepchem/models/tensorgraph/layers.py +225 −122 Original line number Diff line number Diff line Loading @@ -96,13 +96,13 @@ def convert_to_layers(in_layers): raise ValueError("convert_to_layers must be invoked on layers or tensors") return layers class Conv1DLayer(Layer): class Conv1D(Layer): def __init__(self, width, out_channels, **kwargs): self.width = width self.out_channels = out_channels self.out_tensor = None super(Conv1DLayer, self).__init__(**kwargs) super(Conv1D, self).__init__(**kwargs) def create_tensor(self, in_layers=None): if in_layers is None: Loading Loading @@ -190,6 +190,7 @@ class Dense(Layer): class Flatten(Layer): """Flatten every dimension except the first""" def __init__(self, **kwargs): super(Flatten, self).__init__(**kwargs) Loading @@ -216,21 +217,28 @@ class Reshape(Layer): self.shape = shape super(Reshape, self).__init__(**kwargs) def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor self.out_tensor = tf.reshape(parent_tensor, self.shape) return self.out_tensor class Transpose(Layer): def __init__(self, out_shape, **kwargs): def __init__(self, perm, **kwargs): super(Transpose, self).__init__(**kwargs) self.out_shape = out_shape self.perm = perm def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Only One Parent to Transpose over") self.out_tensor = tf.transpose(self.in_layers[0].out_tensor, self.out_shape) self.out_tensor = tf.transpose(in_layers[0].out_tensor, self.perm) return self.out_tensor Loading @@ -239,14 +247,18 @@ class CombineMeanStd(Layer): def __init__(self, **kwargs): super(CombineMeanStd, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: raise ValueError("Must have two self.in_layers") mean_parent, std_parent = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError("Must have two in_layers") mean_parent, std_parent = in_layers[0], in_layers[1] mean_parent_tensor, std_parent_tensor = mean_parent.out_tensor, std_parent.out_tensor sample_noise = tf.random_normal( mean_parent_tensor.get_shape(), 0, 1, dtype=tf.float32) self.out_tensor = mean_parent_tensor + (std_parent_tensor * sample_noise) return self.out_tensor class Repeat(Layer): Loading @@ -255,13 +267,17 @@ class Repeat(Layer): self.n_times = n_times super(Repeat, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor parent_tensor = in_layers[0].out_tensor t = tf.expand_dims(parent_tensor, 1) pattern = tf.stack([1, self.n_times, 1]) self.out_tensor = tf.tile(t, pattern) return self.out_tensor class GRU(Layer): Loading @@ -272,11 +288,14 @@ class GRU(Layer): self.batch_size = batch_size super(GRU, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor gru_cell = tf.nn.rnn_cell.GRUCell(self.n_hidden) parent_tensor = in_layers[0].out_tensor gru_cell = tf.contrib.rnn.GRUCell(self.n_hidden) initial_gru_state = gru_cell.zero_state(self.batch_size, tf.float32) rnn_outputs, rnn_states = tf.nn.dynamic_rnn( gru_cell, Loading @@ -285,20 +304,27 @@ class GRU(Layer): scope=self.name) projection = lambda x: tf.contrib.layers.linear(x, num_outputs=self.out_channels, activation_fn=tf.nn.sigmoid) self.out_tensor = tf.map_fn(projection, rnn_outputs) return self.out_tensor class TimeSeriesDense(Layer): def __init__(self, out_channels, **kwargs): self.out_channels = out_channels super(TimeSeriesDense, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must have one parent") parent_tensor = self.in_layers[0].out_tensor dense_fn = lambda x: tf.contrib.layers.fully_connected(x, num_outputs=self.out_channels, parent_tensor = in_layers[0].out_tensor dense_fn = lambda x: tf.contrib.layers.fully_connected( x, num_outputs=self.out_channels, activation_fn=tf.nn.sigmoid) self.out_tensor = tf.map_fn(dense_fn, parent_tensor) return self.out_tensor class Input(Layer): Loading @@ -309,9 +335,12 @@ class Input(Layer): super(Input, self).__init__(**kwargs) self.op_type = "cpu" def _create_tensor(self): if len(self.in_layers) > 0: queue = self.in_layers[0] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 0: queue = in_layers[0] placeholder = queue.out_tensors[self.get_pre_q_name()] self.out_tensor = tf.placeholder_with_default(placeholder, self.shape) return self.out_tensor Loading Loading @@ -344,13 +373,16 @@ class Weights(Input): super(Weights, self).__init__(**kwargs) class L2LossLayer(Layer): class L2Loss(Layer): def __init__(self, **kwargs): super(L2LossLayer, self).__init__(**kwargs) super(L2Loss, self).__init__(**kwargs) def _create_tensor(self): guess, label = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) guess, label = in_layers[0], in_layers[1] self.out_tensor = tf.reduce_mean( tf.square(guess.out_tensor - label.out_tensor)) return self.out_tensor Loading @@ -361,10 +393,13 @@ class SoftMax(Layer): def __init__(self, **kwargs): super(SoftMax, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Must only Softmax single parent") parent = self.in_layers[0] parent = in_layers[0] self.out_tensor = tf.contrib.layers.softmax(parent.out_tensor) return self.out_tensor Loading @@ -375,11 +410,14 @@ class Concat(Layer): self.axis = axis super(Concat, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) == 1: self.out_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) == 1: self.out_tensor = in_layers[0].out_tensor return self.out_tensor out_tensors = [x.out_tensor for x in self.in_layers] out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.concat(out_tensors, axis=self.axis) return self.out_tensor Loading @@ -394,11 +432,13 @@ class InteratomicL2Distances(Layer): self.ndim = ndim super(InteratomicL2Distances, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError("InteratomicDistances requires coords,nbr_list") coords, nbr_list = (self.in_layers[0].out_tensor, self.in_layers[1].out_tensor) coords, nbr_list = (in_layers[0].out_tensor, in_layers[1].out_tensor) N_atoms, M_nbrs, ndim = self.N_atoms, self.M_nbrs, self.ndim # Shape (N_atoms, M_nbrs, ndim) nbr_coords = tf.gather(coords, nbr_list) Loading @@ -416,10 +456,13 @@ class SoftMaxCrossEntropy(Layer): def __init__(self, **kwargs): super(SoftMaxCrossEntropy, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) != 2: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 2: raise ValueError() labels, logits = self.in_layers[0].out_tensor, self.in_layers[1].out_tensor labels, logits = in_layers[0].out_tensor, in_layers[1].out_tensor self.out_tensor = tf.nn.softmax_cross_entropy_with_logits( logits=logits, labels=labels) self.out_tensor = tf.reshape(self.out_tensor, [-1, 1]) Loading @@ -432,12 +475,15 @@ class ReduceMean(Layer): self.axis = axis super(ReduceMean, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) > 1: out_tensors = [x.out_tensor for x in self.in_layers] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.stack(out_tensors) else: self.out_tensor = self.in_layers[0].out_tensor self.out_tensor = in_layers[0].out_tensor self.out_tensor = tf.reduce_mean(self.out_tensor) return self.out_tensor Loading @@ -445,10 +491,13 @@ class ReduceMean(Layer): class ToFloat(Layer): def _create_tensor(self): if len(self.in_layers) > 1: def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: raise ValueError("Only one layer supported.") self.out_tensor = tf.to_float(self.in_layers[0].out_tensor) self.out_tensor = tf.to_float(in_layers[0].out_tensor) return self.out_tensor Loading @@ -458,12 +507,15 @@ class ReduceSum(Layer): self.axis = axis super(ReduceSum, self).__init__(**kwargs) def _create_tensor(self): if len(self.in_layers) > 1: out_tensors = [x.out_tensor for x in self.in_layers] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) > 1: out_tensors = [x.out_tensor for x in in_layers] self.out_tensor = tf.stack(out_tensors) else: self.out_tensor = self.in_layers[0].out_tensor self.out_tensor = in_layers[0].out_tensor self.out_tensor = tf.reduce_sum(self.out_tensor, axis=self.axis) return self.out_tensor Loading @@ -475,23 +527,29 @@ class ReduceSquareDifference(Layer): self.axis = axis super(ReduceSquareDifference, self).__init__(**kwargs) def _create_tensor(self): a = self.in_layers[0].out_tensor b = self.in_layers[1].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) a = in_layers[0].out_tensor b = in_layers[1].out_tensor self.out_tensor = tf.reduce_mean( tf.squared_difference(a, b), axis=self.axis) return self.out_tensor class Conv2d(Layer): class Conv2D(Layer): def __init__(self, num_outputs, kernel_size=5, **kwargs): self.num_outputs = num_outputs self.kernel_size = kernel_size super(Conv2d, self).__init__(**kwargs) super(Conv2D, self).__init__(**kwargs) def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor out_tensor = tf.contrib.layers.conv2d( parent_tensor, num_outputs=self.num_outputs, Loading @@ -500,6 +558,7 @@ class Conv2d(Layer): activation_fn=tf.nn.relu, normalizer_fn=tf.contrib.layers.batch_norm) self.out_tensor = out_tensor return self.out_tensor class MaxPool(Layer): Loading @@ -514,8 +573,11 @@ class MaxPool(Layer): self.padding = padding super(MaxPool, self).__init__(**kwargs) def _create_tensor(self): in_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) in_tensor = in_layers[0].out_tensor self.out_tensor = tf.nn.max_pool( in_tensor, ksize=self.ksize, strides=self.strides, padding=self.padding) return self.out_tensor Loading @@ -535,12 +597,15 @@ class InputFifoQueue(Layer): super(InputFifoQueue, self).__init__(**kwargs) self.op_type = "cpu" def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if self.dtypes is None: self.dtypes = [tf.float32] * len(self.shapes) self.queue = tf.FIFOQueue( self.capacity, self.dtypes, shapes=self.shapes, names=self.names) feed_dict = {x.name: x.out_tensor for x in self.in_layers} feed_dict = {x.out_tensor.name: x.out_tensor for x in in_layers} self.out_tensor = self.queue.enqueue(feed_dict) self.close_op = self.queue.close() self.out_tensors = self.queue.dequeue() Loading @@ -557,7 +622,7 @@ class InputFifoQueue(Layer): self.queue.close() class GraphConvLayer(Layer): class GraphConv(Layer): def __init__(self, out_channel, Loading @@ -570,11 +635,14 @@ class GraphConvLayer(Layer): self.max_degree = max_deg self.num_deg = 2 * max_deg + (1 - min_deg) self.activation_fn = activation_fn super(GraphConvLayer, self).__init__(**kwargs) super(GraphConv, self).__init__(**kwargs) def _create_tensor(self): # self.in_layers = [atom_features, deg_slice, membership, deg_adj_list placeholders...] in_channels = self.in_layers[0].out_tensor.get_shape()[-1].value def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) # in_layers = [atom_features, deg_slice, membership, deg_adj_list placeholders...] in_channels = in_layers[0].out_tensor.get_shape()[-1].value # Generate the nb_affine weights and biases self.W_list = [ Loading @@ -588,11 +656,11 @@ class GraphConvLayer(Layer): ] # Extract atom_features atom_features = self.in_layers[0].out_tensor atom_features = in_layers[0].out_tensor # Extract graph topology deg_slice = self.in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in self.in_layers[3:]] deg_slice = in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in in_layers[3:]] # Perform the mol conv # atom_features = graph_conv(atom_features, deg_adj_lists, deg_slice, Loading Loading @@ -668,17 +736,21 @@ class GraphConvLayer(Layer): self.out_tensor, self.W_list, self.b_list = tensors class GraphPoolLayer(Layer): class GraphPool(Layer): def __init__(self, min_degree=0, max_degree=10, **kwargs): self.min_degree = min_degree self.max_degree = max_degree super(GraphPoolLayer, self).__init__(**kwargs) super(GraphPool, self).__init__(**kwargs) def _create_tensor(self): atom_features = self.in_layers[0].out_tensor deg_slice = self.in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in self.in_layers[3:]] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) atom_features = in_layers[0].out_tensor deg_slice = in_layers[1].out_tensor deg_adj_lists = [x.out_tensor for x in in_layers[3:]] # Perform the mol gather # atom_features = graph_pool(atom_features, deg_adj_lists, deg_slice, Loading Loading @@ -721,12 +793,16 @@ class GraphGather(Layer): self.activation_fn = activation_fn super(GraphGather, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) # x = [atom_features, deg_slice, membership, deg_adj_list placeholders...] atom_features = self.in_layers[0].out_tensor atom_features = in_layers[0].out_tensor # Extract graph topology membership = self.in_layers[2].out_tensor membership = in_layers[2].out_tensor # Perform the mol gather Loading Loading @@ -757,44 +833,55 @@ class GraphGather(Layer): return mol_features class BatchNormLayer(Layer): class BatchNorm(Layer): def _create_tensor(self): parent_tensor = self.in_layers[0].out_tensor def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) parent_tensor = in_layers[0].out_tensor self.out_tensor = tf.layers.batch_normalization(parent_tensor) return self.out_tensor class WeightedError(Layer): def _create_tensor(self): entropy, weights = self.in_layers[0], self.in_layers[1] def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) entropy, weights = in_layers[0], in_layers[1] self.out_tensor = tf.reduce_sum(entropy.out_tensor * weights.out_tensor) return self.out_tensor class Cutoff(Layer): """Truncates interactions that are too far away.""" def _create_tensor(self): if len(self.in_layers) != 2: raise ValueError("Cutoff must be given distances and energies.") d, x = self.in_layers[0].out_tensor, self.in_layers[1].out_tensor self.out_tensor = tf.where(d < 8, x, tf.zeros_like(x)) return self.out_tensor class VinaFreeEnergy(Layer): """Computes free-energy as defined by Autodock Vina.""" """Computes free-energy as defined by Autodock Vina. def __init__(self, stddev=.3, Nrot=1, **kwargs): TODO(rbharath): Make this layer support batching. """ def __init__(self, N_atoms, M_nbrs, ndim, nbr_cutoff, start, stop, stddev=.3, Nrot=1, **kwargs): self.stddev = stddev # Number of rotatable bonds # TODO(rbharath): Vina actually sets this per-molecule. See if makes # a difference. self.Nrot = Nrot self.N_atoms = N_atoms self.M_nbrs = M_nbrs self.ndim = ndim self.nbr_cutoff = nbr_cutoff self.start = start self.stop = stop super(VinaFreeEnergy, self).__init__(**kwargs) def cutoff(self, d, x): out_tensor = tf.where(d < 8, x, tf.zeros_like(x)) return out_tensor def nonlinearity(self, c): """Computes non-linearity used in Vina.""" w = tf.Variable(tf.random_normal((1,), stddev=self.stddev)) Loading Loading @@ -836,13 +923,13 @@ class VinaFreeEnergy(Layer): out_tensor = tf.exp(-((d - 3) / 2)**2) return out_tensor def _create_tensor(self): def create_tensor(self, in_layers=None): """ Parameters ---------- X: tf.Tensor of shape (B, N, d) X: tf.Tensor of shape (N, d) Coordinates/features. Z: tf.Tensor of shape (B, N) Z: tf.Tensor of shape (N) Atomic numbers of neighbor atoms. Returns Loading @@ -850,16 +937,21 @@ class VinaFreeEnergy(Layer): layer: tf.Tensor of shape (B) The free energy of each complex in batch """ X = self.in_layers[0].out_tensor Z = self.in_layers[2].out_tensor if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) X = in_layers[0].out_tensor Z = in_layers[1].out_tensor # TODO(rbharath): This layer shouldn't be neighbor-listing. Make # neighbors lists an argument instead of a part of this layer. nbr_list = NeighborList( self.N_atoms, self.M_nbrs, self.ndim, self.nbr_cutoff, self.start, self.stop)(coords) self.stop)(X) # Shape (N, M) dists = InteratomicL2Distances( self.N_atoms, self.M_nbrs, self.ndim)(coords, nbr_list) self.N_atoms, self.M_nbrs, self.ndim)(X, nbr_list) repulsion = self.repulsion(dists) hydrophobic = self.hydrophobic(dists) Loading @@ -872,12 +964,12 @@ class VinaFreeEnergy(Layer): repulsion, hydrophobic, hbond, gauss_1, gauss_2) # Shape (N, M) thresholded = Cutoff()(dists, interactions) thresholded = self.cutoff(dists, interactions) free_energies = self.nonlinearity(thresholded) free_energy = ReduceSum()(free_energies) self.output_tensor = free_energy self.out_tensor = free_energy return self.out_tensor Loading @@ -888,10 +980,13 @@ class WeightedLinearCombo(Layer): self.std = std super(WeightedLinearCombo, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) weights = [] out_tensor = None for in_layer in self.in_layers: for in_layer in in_layers: w = tf.Variable(tf.random_normal([ 1, ], stddev=self.std)) Loading @@ -908,6 +1003,8 @@ class NeighborList(Layer): Neighbor-lists (also called Verlet Lists) are a tool for grouping atoms which are close to each other spatially TODO(rbharath): Make this layer support batching. """ def __init__(self, N_atoms, M_nbrs, ndim, nbr_cutoff, start, stop, **kwargs): Loading Loading @@ -935,12 +1032,15 @@ class NeighborList(Layer): self.stop = stop super(NeighborList, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): """Creates tensors associated with neighbor-listing.""" if len(self.in_layers) != 1: if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) if len(in_layers) != 1: raise ValueError("Only One Parent to NeighborList over %s" % self.in_layers) parent = self.in_layers[0] in_layers) parent = in_layers[0] if len(parent.out_tensor.get_shape()) != 2: # TODO(rbharath): Support batching raise ValueError("Parent tensor must be (num_atoms, ndum)") Loading Loading @@ -1209,7 +1309,7 @@ class AtomicConvolution(Layer): self.atom_types = atom_types super(AtomicConvolution, self).__init__(**kwargs) def _create_tensor(self): def create_tensor(self, in_layers=None): """ Parameters ---------- Loading @@ -1225,10 +1325,13 @@ class AtomicConvolution(Layer): layer: tf.Tensor of shape (l, B, N) A new tensor representing the output of the atomic conv layer """ if in_layers is None: in_layers = self.in_layers in_layers = convert_to_layers(in_layers) X = self.in_layers[0].out_tensor Nbrs = tf.to_int32(self.in_layers[1].out_tensor) Nbrs_Z = self.in_layers[2].out_tensor X = in_layers[0].out_tensor Nbrs = tf.to_int32(in_layers[1].out_tensor) Nbrs_Z = in_layers[2].out_tensor # N: Maximum number of atoms # M: Maximum number of neighbors Loading
deepchem/models/tensorgraph/models/graph_conv.py +1 −1 Original line number Diff line number Diff line import tensorflow as tf from deepchem.models.tensorgraph.tensor_graph import TensorGraph from deepchem.models.tensorgraph.layers import Input, Dense, Concat, SoftMax, SoftMaxCrossEntropy, Layer, \ GraphConvLayer, BatchNormLayer, GraphPoolLayer, GraphGather, WeightedError GraphConv, BatchNormLayer, GraphPoolLayer, GraphGather, WeightedError from deepchem.metrics import to_one_hot from deepchem.feat.mol_graphs import ConvMol import time Loading
deepchem/models/tensorgraph/models/graph_models.py +2 −2 Original line number Diff line number Diff line Loading @@ -4,8 +4,8 @@ import six from deepchem.models.tensorgraph.tensor_graph import TensorGraph from deepchem.utils.evaluate import GeneratorEvaluator from deepchem.models.tensorgraph.layers import Input, BatchNormLayer, Dense, \ SoftMax, SoftMaxCrossEntropy, L2LossLayer, Concat, WeightedError, Label, Weights, Feature from deepchem.models.tensorgraph.layers import Input, BatchNorm, Dense, \ SoftMax, SoftMaxCrossEntropy, L2Loss, Concat, WeightedError, Label, Weights, Feature from deepchem.models.tensorgraph.graph_layers import WeaveLayer, WeaveGather, \ Combine_AP, Separate_AP, DTNNEmbedding, DTNNStep, DTNNGather, DAGLayer, DAGGather from deepchem.metrics import to_one_hot, from_one_hot Loading
deepchem/models/tensorgraph/tensor_graph.py +1 −1 Original line number Diff line number Diff line Loading @@ -317,7 +317,7 @@ class TensorGraph(Model): for node in order: with tf.name_scope(node): node_layer = self.layers[node] node_layer._create_tensor() node_layer.create_tensor() self.built = True for layer in self.layers.values(): Loading
deepchem/models/tensorgraph/tests/test_docking.py +0 −19 Original line number Diff line number Diff line Loading @@ -470,22 +470,3 @@ class TestDocking(test_util.TensorFlowTestCase): tg = dc.models.TensorGraph(learning_rate=0.1, use_queue=False) tg.set_loss(loss) tg.fit_generator(databag.iterbatches(epochs=1)) def test_interatomic_distances(self): """Test that the interatomic distance calculation works.""" N_atoms = 5 M_nbrs = 2 ndim = 3 with self.test_session() as sess: coords = np.random.rand(N_atoms, ndim) nbr_list = np.random.randint(0, N_atoms, size=(N_atoms, M_nbrs)) coords_tensor = tf.convert_to_tensor(coords) nbr_list_tensor = tf.convert_to_tensor(nbr_list) dist_tensor = InteratomicL2Distances(N_atoms, M_nbrs, ndim)( coords_tensor, nbr_list_tensor) dists = dist_tensor.eval() assert dists.shape == (N_atoms, M_nbrs)
