dag transformer (a38ee74f) · Commits · 钟慕尧 / deepchem

deepchem/molnet/preset_hyper_parameters.py

+2 −0

Original line number	Diff line number	Diff line
		@@ -63,6 +63,7 @@ hps['dag'] = {
		'nb_epoch': 50,
		'learning_rate': 0.001,
		'n_graph_feat': 30,
		'default_max_atoms': 60,
		'seed': 123
		}
		hps['weave'] = {
		@@ -137,6 +138,7 @@ hps['dag_regression'] = {
		'nb_epoch': 50,
		'learning_rate': 0.001,
		'n_graph_feat': 30,
		'default_max_atoms': 60,
		'seed': 123
		}
		hps['weave_regression'] = {

deepchem/molnet/run_benchmark_models.py

+6 −2

Original line number	Diff line number	Diff line
		@@ -227,12 +227,14 @@ def benchmark_classification(train_dataset,
		nb_epoch = hyper_parameters['nb_epoch']
		learning_rate = hyper_parameters['learning_rate']
		n_graph_feat = hyper_parameters['n_graph_feat']
		default_max_atoms = hyper_parameters['default_max_atoms']

		max_atoms_train = max([mol.get_num_atoms() for mol in train_dataset.X])
		max_atoms_valid = max([mol.get_num_atoms() for mol in valid_dataset.X])
		max_atoms_test = max([mol.get_num_atoms() for mol in test_dataset.X])
		max_atoms = max([max_atoms_train, max_atoms_valid, max_atoms_test])

		max_atoms = min([max_atoms, default_max_atoms])
		print('Maximum number of atoms: %i' % max_atoms)
		reshard_size = 256
		transformer = deepchem.trans.DAGTransformer(max_atoms=max_atoms)
		train_dataset.reshard(reshard_size)
		@@ -560,12 +562,14 @@ def benchmark_regression(train_dataset,
		nb_epoch = hyper_parameters['nb_epoch']
		learning_rate = hyper_parameters['learning_rate']
		n_graph_feat = hyper_parameters['n_graph_feat']
		default_max_atoms = hyper_parameters['default_max_atoms']

		max_atoms_train = max([mol.get_num_atoms() for mol in train_dataset.X])
		max_atoms_valid = max([mol.get_num_atoms() for mol in valid_dataset.X])
		max_atoms_test = max([mol.get_num_atoms() for mol in test_dataset.X])
		max_atoms = max([max_atoms_train, max_atoms_valid, max_atoms_test])

		max_atoms = min([max_atoms, default_max_atoms])
		print('Maximum number of atoms: %i' % max_atoms)
		reshard_size = 512
		transformer = deepchem.trans.DAGTransformer(max_atoms=max_atoms)
		train_dataset.reshard(reshard_size)

deepchem/trans/transformers.py

+10 −5

Original line number	Diff line number	Diff line
		@@ -865,14 +865,14 @@ class DAGTransformer(Transformer):
		# DAG starts from the target atom, calculation should go in reverse
		for edge in reversed(DAG):
		# `edge[1]` is the parent of `edge[0]`
		parent[edge[0]].append(edge[1])
		parent[edge[0]].append(edge[1]%max_atoms)
		# all the parents of `edge[1]` is also the parents of `edge[0]`
		parent[edge[0]].extend(parent[edge[1]])
		# after this loop, `parents[i]` includes all parents of atom i

		for ids, atom in enumerate(parent):
		# manually adding the atom index into its parents list
		parent[ids].insert(0, ids)
		parent[ids].insert(0, ids%max_atoms)
		# after this loop, `parents[i][0]` is i, `parents[i][1:]` are all parents of atom i

		# atoms with less parents(farther from the target atom) come first.
		@@ -885,8 +885,13 @@ class DAGTransformer(Transformer):
		for ids, atom in enumerate(parent):
		n_par = len(atom)
		# padding with `max_atoms`
		if n_par < max_atoms:
		parent[ids].extend([max_atoms for i in range(max_atoms - n_par)])
		if n_par > max_atoms:
		parent[ids] = parent[ids][:max_atoms]

		if len(parent) > max_atoms:
		parent = parent[-max_atoms:]
		while len(parent) < max_atoms:
		# padding
		parent.insert(0, [max_atoms] * max_atoms)

Admin message