Loading deepchem/data/supports.py +41 −0 Original line number Diff line number Diff line Loading @@ -9,6 +9,47 @@ import time import numpy as np from deepchem.data import NumpyDataset def remove_dead_examples(dataset): """Removes compounds with no weight. Parameters ---------- dataset: dc.data.Dataset Source dataset. """ w = dataset.w nonzero_inds = np.nonzero(np.sum(w, axis=1)) # Remove support indices X = dataset.X[nonzero_inds] y = dataset.y[nonzero_inds] w = dataset.w[nonzero_inds] ids = dataset.ids[nonzero_inds] return NumpyDataset(X, y, w, ids) def dataset_difference(dataset, remove): """Removes the compounds in remove from dataset. Parameters ---------- dataset: dc.data.Dataset Source dataset. remove: dc.data.Dataset Dataset whose overlap will be removed. """ remove_ids = set(remove.ids) keep_inds = [ind for ind in range(len(dataset)) if dataset.ids[ind] not in remove_ids] # Remove support indices X = dataset.X[keep_inds] y = dataset.y[keep_inds] w = dataset.w[keep_inds] ids = dataset.ids[keep_inds] return NumpyDataset(X, y, w, ids) def get_task_dataset_minus_support(dataset, support, task): """Gets data for specified task, minus support points. Loading deepchem/data/tests/test_support_generator.py +81 −0 Original line number Diff line number Diff line Loading @@ -19,6 +19,27 @@ class TestSupports(unittest.TestCase): Test that support generation happens properly. """ def test_remove_dead_examples(self): """Tests that examples with zero weight are removed.""" n_samples = 100 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) p = .05 ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.binomial(1, p, size=(n_samples, n_tasks)) w = np.random.binomial(1, p, size=(n_samples, n_tasks)) num_nonzero = np.count_nonzero(np.sum(w, axis=1)) dataset = dc.data.NumpyDataset(X, y, w, ids) cleared_dataset = dc.data.remove_dead_examples(dataset) assert len(cleared_dataset) == num_nonzero def test_get_task_support_simple(self): """Tests that get_task_support samples correctly.""" n_samples = 20 Loading Loading @@ -181,6 +202,34 @@ class TestSupports(unittest.TestCase): np.testing.assert_array_equal(task_dataset.w, w[n_support:]) np.testing.assert_array_equal(task_dataset.ids, ids[n_support:]) def test_dataset_difference_simple(self): """Test that fixed index can be removed from dataset.""" n_samples = 20 n_remove = 5 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.randint(2, size=(n_samples, n_tasks)) w = np.ones((n_samples, n_tasks)) dataset = dc.data.NumpyDataset(X, y, w, ids) remove_dataset = dc.data.NumpyDataset(X[:n_remove], y[:n_remove], w[:n_remove], ids[:n_remove]) out_dataset = dc.data.dataset_difference( dataset, remove_dataset) # Assert all remove elements have been removed assert len(out_dataset) == n_samples - n_remove np.testing.assert_array_equal(out_dataset.X, X[n_remove:]) np.testing.assert_array_equal(out_dataset.y, y[n_remove:]) np.testing.assert_array_equal(out_dataset.w, w[n_remove:]) np.testing.assert_array_equal(out_dataset.ids, ids[n_remove:]) def test_get_task_minus_support(self): """Test that random index support can be removed from dataset.""" n_samples = 10 Loading Loading @@ -212,6 +261,38 @@ class TestSupports(unittest.TestCase): np.testing.assert_array_equal(task_dataset.w, w[data_inds]) np.testing.assert_array_equal(task_dataset.ids, ids[data_inds]) def test_dataset_difference(self): """Test that random index can be removed from dataset.""" n_samples = 10 n_remove = 4 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.randint(2, size=(n_samples, n_tasks)) w = np.ones((n_samples, n_tasks)) dataset = dc.data.NumpyDataset(X, y, w, ids) remove_inds = sorted(np.random.choice( np.arange(n_samples), (n_remove,), replace=False)) remove_dataset = dc.data.NumpyDataset(X[remove_inds], y[remove_inds], w[remove_inds], ids[remove_inds]) out_dataset = dc.data.dataset_difference( dataset, remove_dataset) # Assert all remove elements have been removed data_inds = sorted(list(set(range(n_samples)) - set(remove_inds))) assert len(out_dataset) == n_samples - n_remove np.testing.assert_array_equal(out_dataset.X, X[data_inds]) np.testing.assert_array_equal(out_dataset.y, y[data_inds]) np.testing.assert_array_equal(out_dataset.w, w[data_inds]) np.testing.assert_array_equal(out_dataset.ids, ids[data_inds]) def test_get_task_minus_support_missing(self): """Test that support can be removed from dataset with missing data""" n_samples = 20 Loading examples/low_data/sider_from_tox21_attn_one_fold.py 0 → 100644 +80 −0 Original line number Diff line number Diff line """ Train low-data attn models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import tempfile import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # Depth of attention module max_depth = 3 # number positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) # Apply an attention lstm layer support_model.join(dc.nn.AttnLSTMEmbedding( test_batch_size, support_batch_size, max_depth)) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores) examples/low_data/sider_from_tox21_res_one_fold.py 0 → 100644 +80 −0 Original line number Diff line number Diff line """ Train low-data res models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import tempfile import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # Depth of attention module max_depth = 3 # num positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) # Apply a residual lstm layer support_model.join(dc.nn.ResiLSTMEmbedding( test_batch_size, support_batch_size, max_depth)) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores) examples/low_data/sider_from_tox21_siamese_one_fold.py 0 → 100644 +74 −0 Original line number Diff line number Diff line """ Train low-data siamese models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # num positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 n_steps_per_trial = 1 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores) Loading
deepchem/data/supports.py +41 −0 Original line number Diff line number Diff line Loading @@ -9,6 +9,47 @@ import time import numpy as np from deepchem.data import NumpyDataset def remove_dead_examples(dataset): """Removes compounds with no weight. Parameters ---------- dataset: dc.data.Dataset Source dataset. """ w = dataset.w nonzero_inds = np.nonzero(np.sum(w, axis=1)) # Remove support indices X = dataset.X[nonzero_inds] y = dataset.y[nonzero_inds] w = dataset.w[nonzero_inds] ids = dataset.ids[nonzero_inds] return NumpyDataset(X, y, w, ids) def dataset_difference(dataset, remove): """Removes the compounds in remove from dataset. Parameters ---------- dataset: dc.data.Dataset Source dataset. remove: dc.data.Dataset Dataset whose overlap will be removed. """ remove_ids = set(remove.ids) keep_inds = [ind for ind in range(len(dataset)) if dataset.ids[ind] not in remove_ids] # Remove support indices X = dataset.X[keep_inds] y = dataset.y[keep_inds] w = dataset.w[keep_inds] ids = dataset.ids[keep_inds] return NumpyDataset(X, y, w, ids) def get_task_dataset_minus_support(dataset, support, task): """Gets data for specified task, minus support points. Loading
deepchem/data/tests/test_support_generator.py +81 −0 Original line number Diff line number Diff line Loading @@ -19,6 +19,27 @@ class TestSupports(unittest.TestCase): Test that support generation happens properly. """ def test_remove_dead_examples(self): """Tests that examples with zero weight are removed.""" n_samples = 100 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) p = .05 ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.binomial(1, p, size=(n_samples, n_tasks)) w = np.random.binomial(1, p, size=(n_samples, n_tasks)) num_nonzero = np.count_nonzero(np.sum(w, axis=1)) dataset = dc.data.NumpyDataset(X, y, w, ids) cleared_dataset = dc.data.remove_dead_examples(dataset) assert len(cleared_dataset) == num_nonzero def test_get_task_support_simple(self): """Tests that get_task_support samples correctly.""" n_samples = 20 Loading Loading @@ -181,6 +202,34 @@ class TestSupports(unittest.TestCase): np.testing.assert_array_equal(task_dataset.w, w[n_support:]) np.testing.assert_array_equal(task_dataset.ids, ids[n_support:]) def test_dataset_difference_simple(self): """Test that fixed index can be removed from dataset.""" n_samples = 20 n_remove = 5 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.randint(2, size=(n_samples, n_tasks)) w = np.ones((n_samples, n_tasks)) dataset = dc.data.NumpyDataset(X, y, w, ids) remove_dataset = dc.data.NumpyDataset(X[:n_remove], y[:n_remove], w[:n_remove], ids[:n_remove]) out_dataset = dc.data.dataset_difference( dataset, remove_dataset) # Assert all remove elements have been removed assert len(out_dataset) == n_samples - n_remove np.testing.assert_array_equal(out_dataset.X, X[n_remove:]) np.testing.assert_array_equal(out_dataset.y, y[n_remove:]) np.testing.assert_array_equal(out_dataset.w, w[n_remove:]) np.testing.assert_array_equal(out_dataset.ids, ids[n_remove:]) def test_get_task_minus_support(self): """Test that random index support can be removed from dataset.""" n_samples = 10 Loading Loading @@ -212,6 +261,38 @@ class TestSupports(unittest.TestCase): np.testing.assert_array_equal(task_dataset.w, w[data_inds]) np.testing.assert_array_equal(task_dataset.ids, ids[data_inds]) def test_dataset_difference(self): """Test that random index can be removed from dataset.""" n_samples = 10 n_remove = 4 n_features = 3 n_tasks = 1 # Generate dummy dataset np.random.seed(123) ids = np.arange(n_samples) X = np.random.rand(n_samples, n_features) y = np.random.randint(2, size=(n_samples, n_tasks)) w = np.ones((n_samples, n_tasks)) dataset = dc.data.NumpyDataset(X, y, w, ids) remove_inds = sorted(np.random.choice( np.arange(n_samples), (n_remove,), replace=False)) remove_dataset = dc.data.NumpyDataset(X[remove_inds], y[remove_inds], w[remove_inds], ids[remove_inds]) out_dataset = dc.data.dataset_difference( dataset, remove_dataset) # Assert all remove elements have been removed data_inds = sorted(list(set(range(n_samples)) - set(remove_inds))) assert len(out_dataset) == n_samples - n_remove np.testing.assert_array_equal(out_dataset.X, X[data_inds]) np.testing.assert_array_equal(out_dataset.y, y[data_inds]) np.testing.assert_array_equal(out_dataset.w, w[data_inds]) np.testing.assert_array_equal(out_dataset.ids, ids[data_inds]) def test_get_task_minus_support_missing(self): """Test that support can be removed from dataset with missing data""" n_samples = 20 Loading
examples/low_data/sider_from_tox21_attn_one_fold.py 0 → 100644 +80 −0 Original line number Diff line number Diff line """ Train low-data attn models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import tempfile import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # Depth of attention module max_depth = 3 # number positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) # Apply an attention lstm layer support_model.join(dc.nn.AttnLSTMEmbedding( test_batch_size, support_batch_size, max_depth)) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores)
examples/low_data/sider_from_tox21_res_one_fold.py 0 → 100644 +80 −0 Original line number Diff line number Diff line """ Train low-data res models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import tempfile import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # Depth of attention module max_depth = 3 # num positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) # Apply a residual lstm layer support_model.join(dc.nn.ResiLSTMEmbedding( test_batch_size, support_batch_size, max_depth)) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores)
examples/low_data/sider_from_tox21_siamese_one_fold.py 0 → 100644 +74 −0 Original line number Diff line number Diff line """ Train low-data siamese models on Tox21. Test on SIDER. Test last fold only. """ from __future__ import print_function from __future__ import division from __future__ import unicode_literals import numpy as np import deepchem as dc import tensorflow as tf from datasets import load_sider_convmol from datasets import load_tox21_convmol from datasets import to_numpy_dataset # Number of folds for split K = 4 # num positive/negative ligands n_pos = 10 n_neg = 10 # Set batch sizes for network test_batch_size = 128 support_batch_size = n_pos + n_neg nb_epochs = 1 n_train_trials = 2000 n_eval_trials = 20 n_steps_per_trial = 1 learning_rate = 1e-4 log_every_n_samples = 50 # Number of features on conv-mols n_feat = 71 sider_tasks, sider_dataset, _ = load_sider_convmol() sider_dataset = to_numpy_dataset(sider_dataset) tox21_tasks, tox21_dataset, _ = load_tox21_convmol() tox21_dataset = to_numpy_dataset(tox21_dataset) # Define metric metric = dc.metrics.Metric( dc.metrics.roc_auc_score, verbosity="high", mode="classification") # Train support model on train support_model = dc.nn.SequentialSupportGraph(n_feat) # Add layers support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(128, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.GraphConv(64, activation='relu')) support_model.add(dc.nn.GraphPool()) support_model.add(dc.nn.Dense(128, activation='tanh')) support_model.add_test(dc.nn.GraphGather(test_batch_size, activation='tanh')) support_model.add_support(dc.nn.GraphGather(support_batch_size, activation='tanh')) with tf.Session() as sess: model = dc.models.SupportGraphClassifier( sess, support_model, test_batch_size=test_batch_size, support_batch_size=support_batch_size, learning_rate=learning_rate, verbosity="high") ############################################################ DEBUG print("FIT") ############################################################ DEBUG model.fit(tox21_dataset, nb_epochs=nb_epochs, n_episodes_per_epoch=n_train_trials, n_pos=n_pos, n_neg=n_neg, log_every_n_samples=log_every_n_samples) ############################################################ DEBUG print("EVAL") ############################################################ DEBUG scores = model.evaluate( sider_dataset, metric, n_pos, n_neg, n_trials=n_eval_trials) print("Scores on evaluation dataset") print(scores)
