Loading deepchem/models/tensorgraph/models/symmetry_function_regression.py +102 −11 Original line number Diff line number Diff line Loading @@ -136,7 +136,26 @@ class ANIRegression(TensorGraph): def build_grad(self): self.grad = tf.gradients(self.outputs, self.atom_feats) def compute_grad(self, dataset, batch_size=1): def compute_grad(self, dataset, upper_lim=1): """ Computes a batched gradients given an input dataset. Parameters ---------- dataset: dc.Dataset dataset-like object whose X values will be used to compute gradients from upper_lim: int subset of dataset used. Returns ------- np.array Gradients of the input of shape (max_atoms, 4). Note that it is up to the end user to slice this matrix into the correct shape, since it's very likely the derivatives with respect to the atomic numbers are zero. """ with self._get_tf("Graph").as_default(): if not self.built: self.build() Loading @@ -145,41 +164,110 @@ class ANIRegression(TensorGraph): feed_dict = dict() X = dataset.X flags = np.sign(np.array(X[:batch_size, :, 0])) flags = np.sign(np.array(X[:upper_lim, :, 0])) feed_dict[self.atom_flags] = np.stack([flags]*self.max_atoms, axis=2)*\ np.stack([flags]*self.max_atoms, axis=1) feed_dict[self.atom_numbers] = np.array(X[:batch_size, :, 0], dtype=int) feed_dict[self.atom_feats] = np.array(X[:batch_size, :, :], dtype=float) feed_dict[self.atom_numbers] = np.array(X[:upper_lim, :, 0], dtype=int) feed_dict[self.atom_feats] = np.array(X[:upper_lim, :, :], dtype=float) return self.session.run([self.grad], feed_dict=feed_dict) def pred_one(self, X, atomic_nums): """ Makes an energy prediction for a set of atomic coordinates. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- float Predicted energy. Note that the meaning of the returned value is dependent on the training y-values both in semantics (relative vs absolute) and units (kcal/mol vs Hartrees) """ num_atoms = atomic_nums.shape[0] X = X.reshape((num_atoms, 3)) A = atomic_nums.reshape((atomic_nums.shape[0], 1)) Z = np.zeros((23, 4)) Z = np.zeros((self.max_atoms, 4)) Z[:X.shape[0], 1:X.shape[1]+1] = X Z[:A.shape[0], :A.shape[1]] = A X = Z dd = dc.data.NumpyDataset(np.array(X).reshape((1, 23, 4)), np.array(0), np.array(1)) dd = dc.data.NumpyDataset(np.array(X).reshape((1, self.max_atoms, 4)), np.array(0), np.array(1)) return self.predict(dd)[0] def grad_one(self, X, atomic_nums): y = self.pred_one(X, atomic_nums) """ Computes gradients for that of a single structure. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- np.array derivatives of the same shape and type as input parameter X. """ num_atoms = atomic_nums.shape[0] X = X.reshape((num_atoms, 3)) A = atomic_nums.reshape((atomic_nums.shape[0], 1)) Z = np.zeros((23, 4)) Z = np.zeros((self.max_atoms, 4)) Z[:X.shape[0], 1:X.shape[1]+1] = X Z[:A.shape[0], :A.shape[1]] = A X = Z inp = np.array(X).reshape((1, 23, 4)) dd = dc.data.NumpyDataset(inp, np.array([y]), np.array([1])) inp = np.array(X).reshape((1, self.max_atoms, 4)) dd = dc.data.NumpyDataset(inp, np.array([1]), np.array([1])) res = self.compute_grad(dd)[0][0][0] res = res[:num_atoms, 1:] return res.reshape((num_atoms*3,)) def minimize_structure(self, X, atomic_nums): """ Minimizes a structure, as defined by a set of coordinates and their atomic numbers. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- np.array minimized coordinates of the same shape and type as input parameter X. """ num_atoms = atomic_nums.shape[0] # minimizer_kwargs = { # "jac": self.grad_one, # "args": (atomic_nums,), # "method": "CG", # "options": {"disp": True}, # } # res = scipy.optimize.basinhopping( # self.pred_one, # X, # niter=75, # T=0.015, # ~10kcal/mol # stepsize=0.17, # minimizer_kwargs=minimizer_kwargs) res = scipy.optimize.minimize( self.pred_one, X, Loading @@ -188,6 +276,7 @@ class ANIRegression(TensorGraph): method="BFGS", tol=1e-6, options={'disp': True}) return res.x.reshape((num_atoms, 3)) def build_graph(self): Loading @@ -196,7 +285,9 @@ class ANIRegression(TensorGraph): self.atom_flags = Feature(shape=(None, self.max_atoms, self.max_atoms)) self.atom_feats = Feature(shape=(None, self.max_atoms, 4)) previous_layer = ANIFeat(self.atom_feats) previous_layer = ANIFeat( in_layers=self.atom_feats, max_atoms=self.max_atoms) self.featurized = previous_layer Loading deepchem/models/tensorgraph/models/test_symmetry_functions.py 0 → 100644 +66 −0 Original line number Diff line number Diff line import scipy import numpy as np import unittest from deepchem.models import ANIRegression import deepchem as dc class TestANIRegression(unittest.TestCase): def test_gradients(self): max_atoms = 3 X = np.array([ [1, 5.0, 3.2, 1.1], [6, 1.0, 3.4, -1.1], [1, 2.3, 3.4, 2.2] ]) X = X.reshape((1, X.shape[0], X.shape[1])) y = np.array([2.0]) y = y.reshape((1,1)) layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] model = ANIRegression( 1, max_atoms, layer_structures=layer_structures, atom_number_cases=atom_number_cases, batch_size=1, learning_rate=0.001, use_queue=False, mode="regression") print(X.shape, y.shape) train_dataset = dc.data.NumpyDataset(X, y, n_tasks=1) model.fit(train_dataset, nb_epoch=2, checkpoint_interval=100) new_x = np.array([ -2.0, 1.2, 2.1, 1.3, -6.4, 3.1, -2.5, 2.4, 5.6, ]) new_atomic_nums = np.array([1,1,6]) grad_approx = scipy.optimize.approx_fprime( new_x, model.pred_one, 1e-4, new_atomic_nums) grad_exact = model.grad_one(new_x, new_atomic_nums) print(grad_approx) print(grad_exact) np.testing.assert_array_almost_equal(grad_approx, grad_exact) if __name__ == '__main__': unittest.main() No newline at end of file examples/roitberg/app.py +1 −0 Original line number Diff line number Diff line import numpy as np from flask import request, abort, Flask import flask Loading examples/roitberg/roitberg.py +41 −38 Original line number Diff line number Diff line Loading @@ -27,7 +27,7 @@ def load_roiterberg_ANI(relative=True): hdf5files = [ 'ani_gdb_s01.h5', 'ani_gdb_s02.h5', # 'ani_gdb_s03.h5', 'ani_gdb_s03.h5', # 'ani_gdb_s04.h5', # 'ani_gdb_s05.h5', # 'ani_gdb_s06.h5', Loading Loading @@ -138,14 +138,31 @@ def broadcast(dataset, metadata): if __name__ == "__main__": max_atoms = 23 batch_size = 64 # CHANGED FROM 16 layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] metric = [ dc.metrics.Metric(dc.metrics.mean_absolute_error, mode="regression"), dc.metrics.Metric(dc.metrics.pearson_r2_score, mode="regression") ] model_dir = "/tmp/ani.pkl" if os.path.exists(model_dir): print("Restoring existing model...") model = dc.models.ANIRegression.load_from_dir(model_dir=model_dir) else: print("Fitting new model...") train_valid_dataset, test_dataset, all_groups = load_roiterberg_ANI(relative=True) splitter = dc.splits.RandomGroupSplitter(broadcast(train_valid_dataset, all_groups)) print("performing 1-fold split...") # n_folds = 5 # for fold in range(n_folds): print("Folding once....") train_dataset, valid_dataset = splitter.train_test_split(train_valid_dataset) Loading @@ -160,23 +177,6 @@ if __name__ == "__main__": valid_dataset = transformer.transform(valid_dataset) test_dataset = transformer.transform(test_dataset) max_atoms = 23 batch_size = 64 # CHANGED FROM 16 layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] metric = [ dc.metrics.Metric(dc.metrics.mean_absolute_error, mode="regression"), dc.metrics.Metric(dc.metrics.pearson_r2_score, mode="regression") ] model_dir = "/tmp/ani.pkl" if os.path.exists(model_dir): print("Restoring existing model...") model = dc.models.ANIRegression.load_from_dir(model_dir=model_dir) else: print("Fitting new model...") model = dc.models.ANIRegression( 1, max_atoms, Loading @@ -189,10 +189,12 @@ if __name__ == "__main__": mode="regression") # For production, set nb_epoch to 100+ model.fit(train_dataset, nb_epoch=1, checkpoint_interval=100) for i in range(10): model.fit(train_dataset, nb_epoch=10, checkpoint_interval=100) print("Saving model...") model.save() print("Done.") print("Evaluating model") train_scores = model.evaluate(train_dataset, metric, transformers) Loading @@ -216,13 +218,14 @@ if __name__ == "__main__": atomic_nums = np.array([1, 8, 1]) print("Prediction of a single test set element:") print("Prediction of a single test set structure:") print(model.pred_one(coords, atomic_nums)) print("Gradient of a single test set element:") print("Gradient of a single test set structure:") print(model.grad_one(coords, atomic_nums)) # print("Minimization of a single test set structure:") # print(model.minimize_structure(coords, atomic_nums)) app.webapp.model = model app.webapp.run(host='0.0.0.0', debug=False) Loading
deepchem/models/tensorgraph/models/symmetry_function_regression.py +102 −11 Original line number Diff line number Diff line Loading @@ -136,7 +136,26 @@ class ANIRegression(TensorGraph): def build_grad(self): self.grad = tf.gradients(self.outputs, self.atom_feats) def compute_grad(self, dataset, batch_size=1): def compute_grad(self, dataset, upper_lim=1): """ Computes a batched gradients given an input dataset. Parameters ---------- dataset: dc.Dataset dataset-like object whose X values will be used to compute gradients from upper_lim: int subset of dataset used. Returns ------- np.array Gradients of the input of shape (max_atoms, 4). Note that it is up to the end user to slice this matrix into the correct shape, since it's very likely the derivatives with respect to the atomic numbers are zero. """ with self._get_tf("Graph").as_default(): if not self.built: self.build() Loading @@ -145,41 +164,110 @@ class ANIRegression(TensorGraph): feed_dict = dict() X = dataset.X flags = np.sign(np.array(X[:batch_size, :, 0])) flags = np.sign(np.array(X[:upper_lim, :, 0])) feed_dict[self.atom_flags] = np.stack([flags]*self.max_atoms, axis=2)*\ np.stack([flags]*self.max_atoms, axis=1) feed_dict[self.atom_numbers] = np.array(X[:batch_size, :, 0], dtype=int) feed_dict[self.atom_feats] = np.array(X[:batch_size, :, :], dtype=float) feed_dict[self.atom_numbers] = np.array(X[:upper_lim, :, 0], dtype=int) feed_dict[self.atom_feats] = np.array(X[:upper_lim, :, :], dtype=float) return self.session.run([self.grad], feed_dict=feed_dict) def pred_one(self, X, atomic_nums): """ Makes an energy prediction for a set of atomic coordinates. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- float Predicted energy. Note that the meaning of the returned value is dependent on the training y-values both in semantics (relative vs absolute) and units (kcal/mol vs Hartrees) """ num_atoms = atomic_nums.shape[0] X = X.reshape((num_atoms, 3)) A = atomic_nums.reshape((atomic_nums.shape[0], 1)) Z = np.zeros((23, 4)) Z = np.zeros((self.max_atoms, 4)) Z[:X.shape[0], 1:X.shape[1]+1] = X Z[:A.shape[0], :A.shape[1]] = A X = Z dd = dc.data.NumpyDataset(np.array(X).reshape((1, 23, 4)), np.array(0), np.array(1)) dd = dc.data.NumpyDataset(np.array(X).reshape((1, self.max_atoms, 4)), np.array(0), np.array(1)) return self.predict(dd)[0] def grad_one(self, X, atomic_nums): y = self.pred_one(X, atomic_nums) """ Computes gradients for that of a single structure. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- np.array derivatives of the same shape and type as input parameter X. """ num_atoms = atomic_nums.shape[0] X = X.reshape((num_atoms, 3)) A = atomic_nums.reshape((atomic_nums.shape[0], 1)) Z = np.zeros((23, 4)) Z = np.zeros((self.max_atoms, 4)) Z[:X.shape[0], 1:X.shape[1]+1] = X Z[:A.shape[0], :A.shape[1]] = A X = Z inp = np.array(X).reshape((1, 23, 4)) dd = dc.data.NumpyDataset(inp, np.array([y]), np.array([1])) inp = np.array(X).reshape((1, self.max_atoms, 4)) dd = dc.data.NumpyDataset(inp, np.array([1]), np.array([1])) res = self.compute_grad(dd)[0][0][0] res = res[:num_atoms, 1:] return res.reshape((num_atoms*3,)) def minimize_structure(self, X, atomic_nums): """ Minimizes a structure, as defined by a set of coordinates and their atomic numbers. Parameters ---------- X: np.array numpy array of shape (a, 3) where a <= max_atoms and dtype is float-like atomic_nums: np.array numpy array of shape (a,) where a is the same as that of X. Returns ------- np.array minimized coordinates of the same shape and type as input parameter X. """ num_atoms = atomic_nums.shape[0] # minimizer_kwargs = { # "jac": self.grad_one, # "args": (atomic_nums,), # "method": "CG", # "options": {"disp": True}, # } # res = scipy.optimize.basinhopping( # self.pred_one, # X, # niter=75, # T=0.015, # ~10kcal/mol # stepsize=0.17, # minimizer_kwargs=minimizer_kwargs) res = scipy.optimize.minimize( self.pred_one, X, Loading @@ -188,6 +276,7 @@ class ANIRegression(TensorGraph): method="BFGS", tol=1e-6, options={'disp': True}) return res.x.reshape((num_atoms, 3)) def build_graph(self): Loading @@ -196,7 +285,9 @@ class ANIRegression(TensorGraph): self.atom_flags = Feature(shape=(None, self.max_atoms, self.max_atoms)) self.atom_feats = Feature(shape=(None, self.max_atoms, 4)) previous_layer = ANIFeat(self.atom_feats) previous_layer = ANIFeat( in_layers=self.atom_feats, max_atoms=self.max_atoms) self.featurized = previous_layer Loading
deepchem/models/tensorgraph/models/test_symmetry_functions.py 0 → 100644 +66 −0 Original line number Diff line number Diff line import scipy import numpy as np import unittest from deepchem.models import ANIRegression import deepchem as dc class TestANIRegression(unittest.TestCase): def test_gradients(self): max_atoms = 3 X = np.array([ [1, 5.0, 3.2, 1.1], [6, 1.0, 3.4, -1.1], [1, 2.3, 3.4, 2.2] ]) X = X.reshape((1, X.shape[0], X.shape[1])) y = np.array([2.0]) y = y.reshape((1,1)) layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] model = ANIRegression( 1, max_atoms, layer_structures=layer_structures, atom_number_cases=atom_number_cases, batch_size=1, learning_rate=0.001, use_queue=False, mode="regression") print(X.shape, y.shape) train_dataset = dc.data.NumpyDataset(X, y, n_tasks=1) model.fit(train_dataset, nb_epoch=2, checkpoint_interval=100) new_x = np.array([ -2.0, 1.2, 2.1, 1.3, -6.4, 3.1, -2.5, 2.4, 5.6, ]) new_atomic_nums = np.array([1,1,6]) grad_approx = scipy.optimize.approx_fprime( new_x, model.pred_one, 1e-4, new_atomic_nums) grad_exact = model.grad_one(new_x, new_atomic_nums) print(grad_approx) print(grad_exact) np.testing.assert_array_almost_equal(grad_approx, grad_exact) if __name__ == '__main__': unittest.main() No newline at end of file
examples/roitberg/app.py +1 −0 Original line number Diff line number Diff line import numpy as np from flask import request, abort, Flask import flask Loading
examples/roitberg/roitberg.py +41 −38 Original line number Diff line number Diff line Loading @@ -27,7 +27,7 @@ def load_roiterberg_ANI(relative=True): hdf5files = [ 'ani_gdb_s01.h5', 'ani_gdb_s02.h5', # 'ani_gdb_s03.h5', 'ani_gdb_s03.h5', # 'ani_gdb_s04.h5', # 'ani_gdb_s05.h5', # 'ani_gdb_s06.h5', Loading Loading @@ -138,14 +138,31 @@ def broadcast(dataset, metadata): if __name__ == "__main__": max_atoms = 23 batch_size = 64 # CHANGED FROM 16 layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] metric = [ dc.metrics.Metric(dc.metrics.mean_absolute_error, mode="regression"), dc.metrics.Metric(dc.metrics.pearson_r2_score, mode="regression") ] model_dir = "/tmp/ani.pkl" if os.path.exists(model_dir): print("Restoring existing model...") model = dc.models.ANIRegression.load_from_dir(model_dir=model_dir) else: print("Fitting new model...") train_valid_dataset, test_dataset, all_groups = load_roiterberg_ANI(relative=True) splitter = dc.splits.RandomGroupSplitter(broadcast(train_valid_dataset, all_groups)) print("performing 1-fold split...") # n_folds = 5 # for fold in range(n_folds): print("Folding once....") train_dataset, valid_dataset = splitter.train_test_split(train_valid_dataset) Loading @@ -160,23 +177,6 @@ if __name__ == "__main__": valid_dataset = transformer.transform(valid_dataset) test_dataset = transformer.transform(test_dataset) max_atoms = 23 batch_size = 64 # CHANGED FROM 16 layer_structures = [128, 128, 64] atom_number_cases = [1, 6, 7, 8] metric = [ dc.metrics.Metric(dc.metrics.mean_absolute_error, mode="regression"), dc.metrics.Metric(dc.metrics.pearson_r2_score, mode="regression") ] model_dir = "/tmp/ani.pkl" if os.path.exists(model_dir): print("Restoring existing model...") model = dc.models.ANIRegression.load_from_dir(model_dir=model_dir) else: print("Fitting new model...") model = dc.models.ANIRegression( 1, max_atoms, Loading @@ -189,10 +189,12 @@ if __name__ == "__main__": mode="regression") # For production, set nb_epoch to 100+ model.fit(train_dataset, nb_epoch=1, checkpoint_interval=100) for i in range(10): model.fit(train_dataset, nb_epoch=10, checkpoint_interval=100) print("Saving model...") model.save() print("Done.") print("Evaluating model") train_scores = model.evaluate(train_dataset, metric, transformers) Loading @@ -216,13 +218,14 @@ if __name__ == "__main__": atomic_nums = np.array([1, 8, 1]) print("Prediction of a single test set element:") print("Prediction of a single test set structure:") print(model.pred_one(coords, atomic_nums)) print("Gradient of a single test set element:") print("Gradient of a single test set structure:") print(model.grad_one(coords, atomic_nums)) # print("Minimization of a single test set structure:") # print(model.minimize_structure(coords, atomic_nums)) app.webapp.model = model app.webapp.run(host='0.0.0.0', debug=False)
