def process(self):
data_list = []
indices_train = []
indices_val = []
indices_test = []
infile = open("test_al_50.index", "r")
for line in infile:
indices_test = line.split(",")
indices_test = [int(i) for i in indices_test]
infile = open("val_al_50.index", "r")
for line in infile:
indices_val = line.split(",")
indices_val = [int(i) for i in indices_val]
infile = open("train_al_50.index", "r")
for line in infile:
indices_train = line.split(",")
indices_train = [int(i) for i in indices_train]
targets = dp.get_dataset("alchemy_full", multigregression=True)
tmp1 = targets[indices_train].tolist()
tmp2 = targets[indices_val].tolist()
tmp3 = targets[indices_test].tolist()
targets = tmp1
targets.extend(tmp2)
targets.extend(tmp3)
node_labels = pre.get_all_node_labels_alchem_1(True, True,
indices_train,
indices_val,
indices_test)
edge_labels = pre.get_all_edge_labels_alchem_1(True, True,
indices_train,
indices_val,
indices_test)
matrices = pre.get_all_matrices_1("alchemy_full", indices_train)
matrices.extend(pre.get_all_matrices_1("alchemy_full", indices_val))
matrices.extend(pre.get_all_matrices_1("alchemy_full", indices_test))
for i, m in enumerate(matrices):
data = Data()
data.edge_index = torch.tensor(matrices[i]).t().contiguous()
one_hot = np.eye(6)[node_labels[i]]
data.x = torch.from_numpy(one_hot).to(torch.float)
one_hot = np.eye(4)[edge_labels[i]]
data.edge_attr = torch.from_numpy(one_hot).to(torch.float)
data.y = torch.from_numpy(np.array([targets[i]])).to(torch.float)
print(data.y.size())
data_list.append(data)
data, slices = self.collate(data_list)
torch.save((data, slices), self.processed_paths[0])
def process(self):
data_list = []
indices_train = []
indices_val = []
indices_test = []
infile = open("test.index.txt", "r")
for line in infile:
indices_test = line.split(",")
indices_test = [int(i) for i in indices_test]
infile = open("val.index.txt", "r")
for line in infile:
indices_val = line.split(",")
indices_val = [int(i) for i in indices_val]
infile = open("train.index.txt", "r")
for line in infile:
indices_train = line.split(",")
indices_train = [int(i) for i in indices_train]
dp.get_dataset("ZINC_train")
dp.get_dataset("ZINC_test")
dp.get_dataset("ZINC_val")
node_labels = pre.get_all_node_labels_ZINC_1(True, True, indices_train,
indices_val, indices_test)
edge_labels = pre.get_all_edge_labels_ZINC_1(True, True, indices_train,
indices_val, indices_test)
targets = pre.read_targets("ZINC_train", indices_train)
targets.extend(pre.read_targets("ZINC_val", indices_val))
targets.extend(pre.read_targets("ZINC_test", indices_test))
matrices = pre.get_all_matrices_1("ZINC_train", indices_train)
matrices.extend(pre.get_all_matrices_1("ZINC_val", indices_val))
matrices.extend(pre.get_all_matrices_1("ZINC_test", indices_test))
for i, m in enumerate(matrices):
data = Data()
data.edge_index = torch.tensor(matrices[i]).t().contiguous()
one_hot = np.eye(21)[node_labels[i]]
data.x = torch.from_numpy(one_hot).to(torch.float)
one_hot = np.eye(21)[edge_labels[i]]
data.edge_attr = torch.from_numpy(one_hot).to(torch.float)
data.y = data.y = torch.from_numpy(np.array([targets[i]
])).to(torch.float)
data_list.append(data)
data, slices = self.collate(data_list)
torch.save((data, slices), self.processed_paths[0])
def process(self):
data_list = []
targets = dp.get_dataset("ZINC_full", regression=True)
targets = targets.tolist()
node_labels = pre.get_all_node_labels_1("ZINC_full", True)
edge_labels = pre.get_all_edge_labels_1("ZINC_full")
matrices = pre.get_all_matrices_1("ZINC_full", list(range(0, 249456)))
for i, m in enumerate(matrices):
data = Data()
data.edge_index = torch.tensor(matrices[i]).t().contiguous()
one_hot = np.eye(28)[node_labels[i]]
data.x = torch.from_numpy(one_hot).to(torch.float)
one_hot = np.eye(3)[edge_labels[i]]
data.edge_attr = torch.from_numpy(one_hot).to(torch.float)
data.y = torch.from_numpy(np.array([targets[i]])).to(torch.float)
data_list.append(data)
data, slices = self.collate(data_list)
torch.save((data, slices), self.processed_paths[0])