Loading deepchem/models/tests/test_overfit.py +6 −1 Original line number Diff line number Diff line Loading @@ -709,7 +709,7 @@ class TestOverfitAPI(test_util.TensorFlowTestCase): beta2=.999, verbosity="high") # Fit trained model model.fit(dataset) model.fit(dataset, nb_epoch=30) model.save() # Eval model on train Loading @@ -717,4 +717,9 @@ class TestOverfitAPI(test_util.TensorFlowTestCase): evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity) scores = evaluator.compute_model_performance([classification_metric]) ############################################################ DEBUG print("scores") print(scores) ############################################################ DEBUG assert scores[classification_metric.name] > .9 deepchem/models/tf_keras_models/graph_models.py +25 −39 Original line number Diff line number Diff line Loading @@ -24,7 +24,7 @@ class SequentialGraphModel(object): placeholders from GraphTopology to each graph layer (from keras_layers) added to the network. Non graph layers don't get the extra placeholders. """ def __init__(self, n_atoms, n_feat, batch_size): def __init__(self, n_atoms, n_feat): """ Parameters ---------- Loading @@ -32,21 +32,18 @@ class SequentialGraphModel(object): (Max?) Number of atoms in system. n_feat: int Number of features per atom. batch_size: int Batch size for training models. """ #super(SequentialGraphModel, self).__init__() self.batch_size = batch_size # Create graph topology and x self.graph_topology = GraphTopology(n_atoms, n_feat, self.batch_size) self.graph_topology = GraphTopology(n_atoms, n_feat) self.output = self.graph_topology.get_atom_features_placeholder() # Keep track of the layers self.layers = [] def add(self, layer): """Adds a new layer to model.""" # Update new value of x # For graphical layers, add connectivity placeholders if type(layer).__name__ in ['GraphConv', 'GraphGather', 'GraphPool']: if (len(self.layers) > 0 and hasattr(self.layers[-1], "__name__")): assert (self.layers[-1].__name__ != "GraphGather", Loading @@ -60,26 +57,6 @@ class SequentialGraphModel(object): # Add layer to the layer list self.layers.append(layer) ''' def graph_gather(self, activation='linear'): gather = GraphGather(self.batch_size, activation=activation) self.layers.append(gather) self.output = gather( [self.output] + self.graph_topology.get_topology_placeholders()) ''' ''' def return_container(self, sess): return GraphContainer(sess, input=self.return_inputs(), output=self.return_outputs(), graph_topology=self.graph_topology) ''' def get_batch_size(self): return self.batch_size def get_graph_topology(self): return self.graph_topology Loading @@ -97,18 +74,28 @@ class SequentialGraphModel(object): return self.layers[layer_id] class SequentialSupportGraphModel(object): def __init__(self, n_atom, n_feat, test_batch_size, support_batch_size): self.test_batch_size = test_batch_size self.support_batch_size = support_batch_size """An analog of Keras Sequential model for test/support models.""" def __init__(self, n_test, n_support, n_feat): """ Parameters ---------- n_test: int Number of test atoms. n_support: int Number of support atoms. n_feat: int Number of atomic features. """ self.n_test = n_test self.n_support = n_support # Create graph topology and x self.test_graph_topology = GraphTopology( n_atom, n_feat, test_batch_size, name='test') n_test, n_feat, name='test') self.support_graph_topology = GraphTopology( n_atom, n_feat, support_batch_size, name='support') self.test = self.test_graph_topology.get_nodes() self.support = self.support_graph_topology.get_nodes() n_support, n_feat, name='support') self.test = self.test_graph_topology.get_atom_features_placeholder() self.support = self.support_graph_topology.get_atom_features_placeholder() # Keep track of the layers self.layers = [] Loading @@ -118,10 +105,6 @@ class SequentialSupportGraphModel(object): def add(self, layer): # Add layer to the layer list self.layers.append(layer) ############################################################# DEBUG print("SequentialSupportGraphModel.add()") print(layer) ############################################################# DEBUG # Update new value of x if type(layer).__name__ in ['GraphConv', 'GraphGather', 'GraphPool']: Loading Loading @@ -157,18 +140,21 @@ class SequentialSupportGraphModel(object): self.bool_pre_gather = False ''' def return_container(self, sess): return SupportGraphContainer( sess, input=self.return_inputs(), output=self.return_outputs(), graph_topology_test=self.test_graph_topology, graph_topology_support=self.support_graph_topology) ''' def return_outputs(self): return [self.test] + [self.support] def return_inputs(self): return self.test_graph_topology.get_inputs() + self.support_graph_topology.get_inputs() return (self.test_graph_topology.get_inputs() + self.support_graph_topology.get_inputs()) def get_layer(self, layer_id): return self.layers[layer_id] deepchem/models/tf_keras_models/graph_topology.py +9 −8 Original line number Diff line number Diff line Loading @@ -26,17 +26,19 @@ def merge_dicts(l): class GraphTopology(object): """Manages placeholders associated with batch of graphs and their topology""" def __init__(self, n_atoms, n_feat, batch_size, name='topology', max_deg=6, def __init__(self, n_atoms, n_feat, name='topology', max_deg=6, min_deg=0): """ Note that batch size is not specified in a GraphTopology object. A batch of molecules must be combined into a disconnected graph and fed to topology directly to handle batches. Parameters ---------- n_atoms: int Number of atoms (max) in graphs. n_feat: int Number of features per atom. batch_size: int Number of molecules per batch. name: str, optional Name of this manager. max_deg: int, optional Loading @@ -47,15 +49,17 @@ class GraphTopology(object): self.n_atoms = n_atoms self.n_feat = n_feat self.batch_size = batch_size self.name = name self.max_deg = max_deg self.min_deg = min_deg self.atom_features_placeholder = Input( #tensor=K.placeholder( # shape=(None, self.n_feat), dtype='float32', # name=self.name+'_atom_features')) tensor=K.placeholder( shape=(None, self.n_feat), dtype='float32', shape=(self.n_atoms, self.n_feat), dtype='float32', name=self.name+'_atom_features')) self.deg_adj_lists_placeholders = [ Input(tensor=K.placeholder( Loading Loading @@ -92,9 +96,6 @@ class GraphTopology(object): """ return self.topology def get_batch_size(self): return self.batch_size def get_atom_features_placeholder(self): return self.atom_features_placeholder Loading deepchem/models/tf_keras_models/keras_layers.py +54 −3 Original line number Diff line number Diff line Loading @@ -456,8 +456,29 @@ class GraphPool(Layer): return atom_features class AttnLSTMEmbedding(Layer): """Implements AttnLSTM as in matching networks paper. References: Matching Networks for One Shot Learning https://arxiv.org/pdf/1606.04080v1.pdf Order Matters: Sequence to sequence for sets https://arxiv.org/abs/1511.06391 """ def __init__(self, max_depth, init='glorot_uniform', activation='linear', dropout=None, **kwargs): """ Parameters ---------- max_depth: int Number of "processing steps" used by sequence-to-sequence for sets model. init: str, optional Type of initialization of weights activation: str, optional Activation for layers. dropout: float, optional Dropout probability """ super(AttnLSTMEmbedding, self).__init__(**kwargs) self.init = initializations.get(init) # Set weight initialization Loading @@ -466,7 +487,9 @@ class AttnLSTMEmbedding(Layer): def build(self, input_shape): print(input_shape) """Initializes trainable weights.""" # x_input_shape = (N_test, N_feat) # xp_input_shape = (N_support, N_feat) x_input_shape, xp_input_shape = input_shape #Unpack N_test = x_input_shape[0] Loading @@ -481,13 +504,41 @@ class AttnLSTMEmbedding(Layer): self.trainable_weights = [self.q_init, self.r_init] def get_output_shape_for(self, input_shape): """Returns the output shape. Same as input_shape. Parameters ---------- input_shape: list Will be of form [(n_test, n_feat), (n_support, n_feat)] Returns ------- list Of same shape as input [(n_test, n_feat), (n_support, n_feat)] """ x_input_shape, xp_input_shape = input_shape #Unpack return input_shape def call(self, x_xp, mask=None): x, xp = x_xp #Unpack print('unpacked') """Execute this layer on input tensors. Parameters ---------- x_xp: list List of two tensors (X, Xp). X should be of shape (n_test, n_feat) and Xp should be of shape (n_support, n_feat) where n_test is the size of the test set, n_support that of the support set, and n_feat is the number of per-atom features. Returns ------- list Returns two tensors of same shape as input. Namely the output shape will be [(n_test, n_feat), (n_support, n_feat)] """ # x is test set, xp is support set. x, xp = x_xp # Get initializations q = self.q_init Loading deepchem/models/tf_keras_models/multitask_classifier.py +4 −4 Original line number Diff line number Diff line Loading @@ -45,7 +45,7 @@ def get_loss_fn(final_loss): class MultitaskGraphClassifier(Model): def __init__(self, sess, model, n_tasks, logdir, def __init__(self, sess, model, n_tasks, logdir, batch_size=50, final_loss='cross_entropy', learning_rate=.001, optimizer_type="adam", learning_rate_decay_time=1000, beta1=.9, beta2=.999, verbosity=None): Loading @@ -58,7 +58,7 @@ class MultitaskGraphClassifier(Model): self.logdir = logdir # Extract model info self.batch_size = self.model.get_batch_size() self.batch_size = batch_size # Get graph topology for x self.graph_topology = self.model.get_graph_topology() self.feat_dim = self.model.get_num_output_features() Loading Loading @@ -180,7 +180,7 @@ class MultitaskGraphClassifier(Model): softmax.append(tf.nn.softmax(logits, name='softmax_%d' % i)) return softmax def fit(self, dataset, nb_epoch=10, batch_size=50, pad_batches=False, def fit(self, dataset, nb_epoch=10, pad_batches=False, max_checkpoints_to_keep=5, log_every_N_batches=50, **kwargs): # Perform the optimization log("Training for %d epochs" % nb_epoch, self.verbosity) Loading @@ -195,7 +195,7 @@ class MultitaskGraphClassifier(Model): log("Starting epoch %d" % epoch, self.verbosity) # ToDo(hraut->rbharath) : what is the ids_b for? Is it the zero's? for batch_num, (X_b, y_b, w_b, ids_b) in enumerate(dataset.iterbatches( batch_size, pad_batches=pad_batches)): self.batch_size, pad_batches=pad_batches)): if batch_num % log_every_N_batches == 0: log("On batch %d" % batch_num, self.verbosity) self.sess.run( Loading Loading
deepchem/models/tests/test_overfit.py +6 −1 Original line number Diff line number Diff line Loading @@ -709,7 +709,7 @@ class TestOverfitAPI(test_util.TensorFlowTestCase): beta2=.999, verbosity="high") # Fit trained model model.fit(dataset) model.fit(dataset, nb_epoch=30) model.save() # Eval model on train Loading @@ -717,4 +717,9 @@ class TestOverfitAPI(test_util.TensorFlowTestCase): evaluator = Evaluator(model, dataset, transformers, verbosity=verbosity) scores = evaluator.compute_model_performance([classification_metric]) ############################################################ DEBUG print("scores") print(scores) ############################################################ DEBUG assert scores[classification_metric.name] > .9
deepchem/models/tf_keras_models/graph_models.py +25 −39 Original line number Diff line number Diff line Loading @@ -24,7 +24,7 @@ class SequentialGraphModel(object): placeholders from GraphTopology to each graph layer (from keras_layers) added to the network. Non graph layers don't get the extra placeholders. """ def __init__(self, n_atoms, n_feat, batch_size): def __init__(self, n_atoms, n_feat): """ Parameters ---------- Loading @@ -32,21 +32,18 @@ class SequentialGraphModel(object): (Max?) Number of atoms in system. n_feat: int Number of features per atom. batch_size: int Batch size for training models. """ #super(SequentialGraphModel, self).__init__() self.batch_size = batch_size # Create graph topology and x self.graph_topology = GraphTopology(n_atoms, n_feat, self.batch_size) self.graph_topology = GraphTopology(n_atoms, n_feat) self.output = self.graph_topology.get_atom_features_placeholder() # Keep track of the layers self.layers = [] def add(self, layer): """Adds a new layer to model.""" # Update new value of x # For graphical layers, add connectivity placeholders if type(layer).__name__ in ['GraphConv', 'GraphGather', 'GraphPool']: if (len(self.layers) > 0 and hasattr(self.layers[-1], "__name__")): assert (self.layers[-1].__name__ != "GraphGather", Loading @@ -60,26 +57,6 @@ class SequentialGraphModel(object): # Add layer to the layer list self.layers.append(layer) ''' def graph_gather(self, activation='linear'): gather = GraphGather(self.batch_size, activation=activation) self.layers.append(gather) self.output = gather( [self.output] + self.graph_topology.get_topology_placeholders()) ''' ''' def return_container(self, sess): return GraphContainer(sess, input=self.return_inputs(), output=self.return_outputs(), graph_topology=self.graph_topology) ''' def get_batch_size(self): return self.batch_size def get_graph_topology(self): return self.graph_topology Loading @@ -97,18 +74,28 @@ class SequentialGraphModel(object): return self.layers[layer_id] class SequentialSupportGraphModel(object): def __init__(self, n_atom, n_feat, test_batch_size, support_batch_size): self.test_batch_size = test_batch_size self.support_batch_size = support_batch_size """An analog of Keras Sequential model for test/support models.""" def __init__(self, n_test, n_support, n_feat): """ Parameters ---------- n_test: int Number of test atoms. n_support: int Number of support atoms. n_feat: int Number of atomic features. """ self.n_test = n_test self.n_support = n_support # Create graph topology and x self.test_graph_topology = GraphTopology( n_atom, n_feat, test_batch_size, name='test') n_test, n_feat, name='test') self.support_graph_topology = GraphTopology( n_atom, n_feat, support_batch_size, name='support') self.test = self.test_graph_topology.get_nodes() self.support = self.support_graph_topology.get_nodes() n_support, n_feat, name='support') self.test = self.test_graph_topology.get_atom_features_placeholder() self.support = self.support_graph_topology.get_atom_features_placeholder() # Keep track of the layers self.layers = [] Loading @@ -118,10 +105,6 @@ class SequentialSupportGraphModel(object): def add(self, layer): # Add layer to the layer list self.layers.append(layer) ############################################################# DEBUG print("SequentialSupportGraphModel.add()") print(layer) ############################################################# DEBUG # Update new value of x if type(layer).__name__ in ['GraphConv', 'GraphGather', 'GraphPool']: Loading Loading @@ -157,18 +140,21 @@ class SequentialSupportGraphModel(object): self.bool_pre_gather = False ''' def return_container(self, sess): return SupportGraphContainer( sess, input=self.return_inputs(), output=self.return_outputs(), graph_topology_test=self.test_graph_topology, graph_topology_support=self.support_graph_topology) ''' def return_outputs(self): return [self.test] + [self.support] def return_inputs(self): return self.test_graph_topology.get_inputs() + self.support_graph_topology.get_inputs() return (self.test_graph_topology.get_inputs() + self.support_graph_topology.get_inputs()) def get_layer(self, layer_id): return self.layers[layer_id]
deepchem/models/tf_keras_models/graph_topology.py +9 −8 Original line number Diff line number Diff line Loading @@ -26,17 +26,19 @@ def merge_dicts(l): class GraphTopology(object): """Manages placeholders associated with batch of graphs and their topology""" def __init__(self, n_atoms, n_feat, batch_size, name='topology', max_deg=6, def __init__(self, n_atoms, n_feat, name='topology', max_deg=6, min_deg=0): """ Note that batch size is not specified in a GraphTopology object. A batch of molecules must be combined into a disconnected graph and fed to topology directly to handle batches. Parameters ---------- n_atoms: int Number of atoms (max) in graphs. n_feat: int Number of features per atom. batch_size: int Number of molecules per batch. name: str, optional Name of this manager. max_deg: int, optional Loading @@ -47,15 +49,17 @@ class GraphTopology(object): self.n_atoms = n_atoms self.n_feat = n_feat self.batch_size = batch_size self.name = name self.max_deg = max_deg self.min_deg = min_deg self.atom_features_placeholder = Input( #tensor=K.placeholder( # shape=(None, self.n_feat), dtype='float32', # name=self.name+'_atom_features')) tensor=K.placeholder( shape=(None, self.n_feat), dtype='float32', shape=(self.n_atoms, self.n_feat), dtype='float32', name=self.name+'_atom_features')) self.deg_adj_lists_placeholders = [ Input(tensor=K.placeholder( Loading Loading @@ -92,9 +96,6 @@ class GraphTopology(object): """ return self.topology def get_batch_size(self): return self.batch_size def get_atom_features_placeholder(self): return self.atom_features_placeholder Loading
deepchem/models/tf_keras_models/keras_layers.py +54 −3 Original line number Diff line number Diff line Loading @@ -456,8 +456,29 @@ class GraphPool(Layer): return atom_features class AttnLSTMEmbedding(Layer): """Implements AttnLSTM as in matching networks paper. References: Matching Networks for One Shot Learning https://arxiv.org/pdf/1606.04080v1.pdf Order Matters: Sequence to sequence for sets https://arxiv.org/abs/1511.06391 """ def __init__(self, max_depth, init='glorot_uniform', activation='linear', dropout=None, **kwargs): """ Parameters ---------- max_depth: int Number of "processing steps" used by sequence-to-sequence for sets model. init: str, optional Type of initialization of weights activation: str, optional Activation for layers. dropout: float, optional Dropout probability """ super(AttnLSTMEmbedding, self).__init__(**kwargs) self.init = initializations.get(init) # Set weight initialization Loading @@ -466,7 +487,9 @@ class AttnLSTMEmbedding(Layer): def build(self, input_shape): print(input_shape) """Initializes trainable weights.""" # x_input_shape = (N_test, N_feat) # xp_input_shape = (N_support, N_feat) x_input_shape, xp_input_shape = input_shape #Unpack N_test = x_input_shape[0] Loading @@ -481,13 +504,41 @@ class AttnLSTMEmbedding(Layer): self.trainable_weights = [self.q_init, self.r_init] def get_output_shape_for(self, input_shape): """Returns the output shape. Same as input_shape. Parameters ---------- input_shape: list Will be of form [(n_test, n_feat), (n_support, n_feat)] Returns ------- list Of same shape as input [(n_test, n_feat), (n_support, n_feat)] """ x_input_shape, xp_input_shape = input_shape #Unpack return input_shape def call(self, x_xp, mask=None): x, xp = x_xp #Unpack print('unpacked') """Execute this layer on input tensors. Parameters ---------- x_xp: list List of two tensors (X, Xp). X should be of shape (n_test, n_feat) and Xp should be of shape (n_support, n_feat) where n_test is the size of the test set, n_support that of the support set, and n_feat is the number of per-atom features. Returns ------- list Returns two tensors of same shape as input. Namely the output shape will be [(n_test, n_feat), (n_support, n_feat)] """ # x is test set, xp is support set. x, xp = x_xp # Get initializations q = self.q_init Loading
deepchem/models/tf_keras_models/multitask_classifier.py +4 −4 Original line number Diff line number Diff line Loading @@ -45,7 +45,7 @@ def get_loss_fn(final_loss): class MultitaskGraphClassifier(Model): def __init__(self, sess, model, n_tasks, logdir, def __init__(self, sess, model, n_tasks, logdir, batch_size=50, final_loss='cross_entropy', learning_rate=.001, optimizer_type="adam", learning_rate_decay_time=1000, beta1=.9, beta2=.999, verbosity=None): Loading @@ -58,7 +58,7 @@ class MultitaskGraphClassifier(Model): self.logdir = logdir # Extract model info self.batch_size = self.model.get_batch_size() self.batch_size = batch_size # Get graph topology for x self.graph_topology = self.model.get_graph_topology() self.feat_dim = self.model.get_num_output_features() Loading Loading @@ -180,7 +180,7 @@ class MultitaskGraphClassifier(Model): softmax.append(tf.nn.softmax(logits, name='softmax_%d' % i)) return softmax def fit(self, dataset, nb_epoch=10, batch_size=50, pad_batches=False, def fit(self, dataset, nb_epoch=10, pad_batches=False, max_checkpoints_to_keep=5, log_every_N_batches=50, **kwargs): # Perform the optimization log("Training for %d epochs" % nb_epoch, self.verbosity) Loading @@ -195,7 +195,7 @@ class MultitaskGraphClassifier(Model): log("Starting epoch %d" % epoch, self.verbosity) # ToDo(hraut->rbharath) : what is the ids_b for? Is it the zero's? for batch_num, (X_b, y_b, w_b, ids_b) in enumerate(dataset.iterbatches( batch_size, pad_batches=pad_batches)): self.batch_size, pad_batches=pad_batches)): if batch_num % log_every_N_batches == 0: log("On batch %d" % batch_num, self.verbosity) self.sess.run( Loading
