import os.path as osp
import numpy as np
import torch
from cogdl.data import Graph, Dataset
from cogdl.utils import download_url
[docs]class BidirectionalOneShotIterator(object):
def __init__(self, dataloader_head, dataloader_tail):
self.iterator_head = self.one_shot_iterator(dataloader_head)
self.iterator_tail = self.one_shot_iterator(dataloader_tail)
self.step = 0
def __next__(self):
self.step += 1
if self.step % 2 == 0:
data = next(self.iterator_head)
else:
data = next(self.iterator_tail)
return data
[docs] @staticmethod
def one_shot_iterator(dataloader):
"""
Transform a PyTorch Dataloader into python iterator
"""
while True:
for data in dataloader:
yield data
[docs]class TestDataset(torch.utils.data.Dataset):
def __init__(self, triples, all_true_triples, nentity, nrelation, mode):
self.len = len(triples)
self.triple_set = set(all_true_triples)
self.triples = triples
self.nentity = nentity
self.nrelation = nrelation
self.mode = mode
def __len__(self):
return self.len
def __getitem__(self, idx):
head, relation, tail = self.triples[idx]
if self.mode == "head-batch":
tmp = [
(0, rand_head) if (rand_head, relation, tail) not in self.triple_set else (-1, head)
for rand_head in range(self.nentity)
]
tmp[head] = (0, head)
elif self.mode == "tail-batch":
tmp = [
(0, rand_tail) if (head, relation, rand_tail) not in self.triple_set else (-1, tail)
for rand_tail in range(self.nentity)
]
tmp[tail] = (0, tail)
else:
raise ValueError("negative batch mode %s not supported" % self.mode)
tmp = torch.LongTensor(tmp)
filter_bias = tmp[:, 0].float()
negative_sample = tmp[:, 1]
positive_sample = torch.LongTensor((head, relation, tail))
return positive_sample, negative_sample, filter_bias, self.mode
[docs] @staticmethod
def collate_fn(data):
positive_sample = torch.stack([_[0] for _ in data], dim=0)
negative_sample = torch.stack([_[1] for _ in data], dim=0)
filter_bias = torch.stack([_[2] for _ in data], dim=0)
mode = data[0][3]
return positive_sample, negative_sample, filter_bias, mode
[docs]class TrainDataset(torch.utils.data.Dataset):
def __init__(self, triples, nentity, nrelation, negative_sample_size, mode):
self.len = len(triples)
self.triples = triples
self.triple_set = set(triples)
self.nentity = nentity
self.nrelation = nrelation
self.negative_sample_size = negative_sample_size
self.mode = mode
self.count = self.count_frequency(triples)
self.true_head, self.true_tail = self.get_true_head_and_tail(self.triples)
def __len__(self):
return self.len
def __getitem__(self, idx):
positive_sample = self.triples[idx]
head, relation, tail = positive_sample
subsampling_weight = self.count[(head, relation)] + self.count[(tail, -relation - 1)]
subsampling_weight = torch.sqrt(1 / torch.Tensor([subsampling_weight]))
negative_sample_list = []
negative_sample_size = 0
while negative_sample_size < self.negative_sample_size:
negative_sample = np.random.randint(self.nentity, size=self.negative_sample_size * 2)
if self.mode == "head-batch":
mask = np.in1d(negative_sample, self.true_head[(relation, tail)], assume_unique=True, invert=True)
elif self.mode == "tail-batch":
mask = np.in1d(negative_sample, self.true_tail[(head, relation)], assume_unique=True, invert=True)
else:
raise ValueError("Training batch mode %s not supported" % self.mode)
negative_sample = negative_sample[mask]
negative_sample_list.append(negative_sample)
negative_sample_size += negative_sample.size
negative_sample = np.concatenate(negative_sample_list)[: self.negative_sample_size]
negative_sample = torch.LongTensor(negative_sample)
positive_sample = torch.LongTensor(positive_sample)
return positive_sample, negative_sample, subsampling_weight, self.mode
[docs] @staticmethod
def collate_fn(data):
positive_sample = torch.stack([_[0] for _ in data], dim=0)
negative_sample = torch.stack([_[1] for _ in data], dim=0)
subsample_weight = torch.cat([_[2] for _ in data], dim=0)
mode = data[0][3]
return positive_sample, negative_sample, subsample_weight, mode
[docs] @staticmethod
def count_frequency(triples, start=4):
"""
Get frequency of a partial triple like (head, relation) or (relation, tail)
The frequency will be used for subsampling like word2vec
"""
count = {}
for head, relation, tail in triples:
if (head, relation) not in count:
count[(head, relation)] = start
else:
count[(head, relation)] += 1
if (tail, -relation - 1) not in count:
count[(tail, -relation - 1)] = start
else:
count[(tail, -relation - 1)] += 1
return count
[docs] @staticmethod
def get_true_head_and_tail(triples):
"""
Build a dictionary of true triples that will
be used to filter these true triples for negative sampling
"""
true_head = {}
true_tail = {}
for head, relation, tail in triples:
if (head, relation) not in true_tail:
true_tail[(head, relation)] = []
true_tail[(head, relation)].append(tail)
if (relation, tail) not in true_head:
true_head[(relation, tail)] = []
true_head[(relation, tail)].append(head)
for relation, tail in true_head:
true_head[(relation, tail)] = np.array(list(set(true_head[(relation, tail)])))
for head, relation in true_tail:
true_tail[(head, relation)] = np.array(list(set(true_tail[(head, relation)])))
return true_head, true_tail
[docs]def read_triplet_data(folder):
filenames = ["train2id.txt", "valid2id.txt", "test2id.txt"]
count = 0
edge_index = []
edge_attr = []
count_list = []
triples = []
num_entities = 0
num_relations = 0
entity_dic = {}
relation_dic = {}
for filename in filenames:
with open(osp.join(folder, filename), "r") as f:
if "train" in filename:
train_start_idx = len(triples)
elif "valid" in filename:
valid_start_idx = len(triples)
elif "test" in filename:
test_start_idx = len(triples)
for line in f:
items = line.strip().split()
if len(items) != 3:
continue
edge_index.append([int(items[0]), int(items[1])])
edge_attr.append(int(items[2]))
triples.append((int(items[0]), int(items[2]), int(items[1])))
if items[0] not in entity_dic:
entity_dic[items[0]] = num_entities
num_entities += 1
if items[1] not in entity_dic:
entity_dic[items[1]] = num_entities
num_entities += 1
if items[2] not in relation_dic:
relation_dic[items[2]] = num_relations
num_relations += 1
count += 1
count_list.append(count)
edge_index = torch.LongTensor(edge_index).t()
edge_attr = torch.LongTensor(edge_attr)
data = Graph()
data.edge_index = edge_index
data.edge_attr = edge_attr
def generate_mask(start, end):
mask = torch.BoolTensor(count)
mask[:] = False
mask[start:end] = True
return mask
data.train_mask = generate_mask(0, count_list[0])
data.val_mask = generate_mask(count_list[0], count_list[1])
data.test_mask = generate_mask(count_list[1], count_list[2])
return data, triples, train_start_idx, valid_start_idx, test_start_idx, num_entities, num_relations
[docs]class KnowledgeGraphDataset(Dataset):
url = "https://cloud.tsinghua.edu.cn/d/d1c733373b014efab986/files/?p=%2F{}%2F{}&dl=1"
def __init__(self, root, name):
self.name = name
super(KnowledgeGraphDataset, self).__init__(root)
self.data = torch.load(self.processed_paths[0])
triple_config = torch.load(self.processed_paths[1])
self.triples = triple_config["triples"]
self._train_start_index = triple_config["train_start_index"]
self._valid_start_index = triple_config["valid_start_index"]
self._test_start_index = triple_config["test_start_index"]
self._num_entities = triple_config["num_entities"]
self._num_relations = triple_config["num_relations"]
@property
def raw_file_names(self):
names = ["train2id.txt", "valid2id.txt", "test2id.txt"]
return names
@property
def processed_file_names(self):
return ["data.pt", "triple_config.pt"]
@property
def train_start_idx(self):
return self._train_start_index
@property
def valid_start_idx(self):
return self._valid_start_index
@property
def test_start_idx(self):
return self._test_start_index
@property
def num_entities(self):
return self._num_entities
@property
def num_relations(self):
return self._num_relations
[docs] def get(self, idx):
assert idx == 0
return self.data
[docs] def download(self):
for name in self.raw_file_names:
download_url(self.url.format(self.name, name), self.raw_dir, name=name)
[docs] def process(self):
(
data,
triples,
train_start_index,
valid_start_index,
test_start_index,
num_entities,
num_relations,
) = read_triplet_data(self.raw_dir)
torch.save(data, self.processed_paths[0])
triple_config = {
"triples": triples,
"train_start_index": train_start_index,
"valid_start_index": valid_start_index,
"test_start_index": test_start_index,
"num_entities": num_entities,
"num_relations": num_relations,
}
torch.save(triple_config, self.processed_paths[1])
[docs]class FB13Datset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "FB13"
path = osp.join(data_path, dataset)
super(FB13Datset, self).__init__(path, dataset)
[docs]class FB15kDatset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "FB15K"
path = osp.join(data_path, dataset)
super(FB15kDatset, self).__init__(path, dataset)
[docs]class FB15k237Datset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "FB15K237"
path = osp.join(data_path, dataset)
super(FB15k237Datset, self).__init__(path, dataset)
[docs]class WN18Datset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "WN18"
path = osp.join(data_path, dataset)
super(WN18Datset, self).__init__(path, dataset)
[docs]class WN18RRDataset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "WN18RR"
path = osp.join(data_path, dataset)
super(WN18RRDataset, self).__init__(path, dataset)
[docs]class FB13SDatset(KnowledgeGraphDataset):
def __init__(self, data_path="data"):
dataset = "FB13S"
path = osp.join(data_path, dataset)
super(FB13SDatset, self).__init__(path, dataset)