-
Notifications
You must be signed in to change notification settings - Fork 4
/
utils.py
288 lines (250 loc) · 9.57 KB
/
utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
import json
from datetime import datetime
import logging
import os
import random
from typing import List
from scipy import cluster
from torch.nn.functional import one_hot
from tqdm import tqdm
import numpy as np
import torch
from sklearn.metrics import precision_score, recall_score, f1_score, accuracy_score
from data import PromptEMData
from args import PromptEMArgs
from summarize import Summarizer
from openprompt import PromptForClassification
import csv
def set_seed(seed):
if seed != -1:
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.deterministic = True
os.environ['PYTHONHASHSEED'] = str(seed)
def set_logger(name):
cur_time = '_' + datetime.now().strftime('%F %T')
name += cur_time
name = name.replace(":", "_")
"""
Write logs to checkpoints and console.
"""
log_file = os.path.join('./logs', name)
logging.basicConfig(
format='%(asctime)s %(levelname)-8s %(message)s',
level=logging.INFO,
datefmt='%Y-%m-%d %H:%M:%S',
filename=log_file,
filemode='w'
)
console = logging.StreamHandler()
console.setLevel(logging.INFO)
formatter = logging.Formatter('%(asctime)s %(levelname)-8s %(message)s')
console.setFormatter(formatter)
logging.getLogger('').addHandler(console)
def evaluate(y_truth, y_pred, return_acc=False):
precision = precision_score(y_truth, y_pred, zero_division=0)
recall = recall_score(y_truth, y_pred, zero_division=0)
f1 = f1_score(y_truth, y_pred, zero_division=0)
acc = accuracy_score(y_truth, y_pred)
if return_acc:
return precision, recall, f1, acc
else:
return precision, recall, f1
def read_ground_truth(file_path, files=None):
if files is None:
files = ["train", "valid", "test"]
x = []
y_truth = []
for file in files:
with open(os.path.join(file_path, f"{file}.csv"), "r") as rd:
for i, line in enumerate(rd.readlines()):
values = line.strip().split(',')
x.append((int(values[0]), int(values[1])))
y_truth.append(int(values[2]))
return x, y_truth
def rel_serialize(cols: List[str], vals: List[str], skip=True, add_token=True) -> str:
sen = ""
for (col, val) in zip(cols, vals):
if skip and val == "":
continue
if col.lower() == "id":
continue
if add_token:
sen += f"COL {col} VAL {val} "
else:
sen += f"{col} {val} "
return sen
def semi_serialize(line: dict, skip=True, add_token=True) -> str:
sen = ""
for (key, val) in line.items():
if key == "id":
continue
if type(val).__name__ == "list":
if skip and len(val) == 0:
continue
if add_token:
sen += f"COL {key} VAL {' '.join(list(map(str, val)))} "
else:
sen += f"{key} {' '.join(list(map(str, val)))} "
elif type(val).__name__ == "dict":
if add_token:
sen += f"COL {key} VAL {semi_serialize(val, skip, add_token)} "
else:
sen += f"{key} {semi_serialize(val, skip, add_token)} "
else:
val = str(val)
if skip and val == "":
continue
if add_token:
sen += f"COL {key} VAL {val} "
else:
sen += f"{key} {val} "
return sen
def read_rel_entities(file_path: str, add_token=True, summarize=False):
entities = []
file_path += ".csv"
with open(file_path, "r") as rd:
data = list(csv.reader(rd))
cols = data[0]
for vals in tqdm(data[1:], desc="read relation entity..."):
entities.append(rel_serialize(cols, vals, add_token=add_token))
if summarize:
summarizer = Summarizer(entities, "roberta-base")
new_entities = []
for ent in tqdm(entities, desc="summarizing..."):
new_entities.append(summarizer.transform_sentence(ent))
entities = new_entities
return entities
def read_semi_entities(file_path: str, skip=True, add_token=True, summarize=False):
entities = []
file_path += ".json"
with open(file_path, "r") as rd:
lines = json.load(rd)
for line in tqdm(lines, desc="read semi entity..."):
entities.append(semi_serialize(line, skip, add_token=add_token))
if summarize:
summarizer = Summarizer(entities, "roberta-base")
new_entities = []
for ent in tqdm(entities, desc="summarizing..."):
new_entities.append(summarizer.transform_sentence(ent))
entities = new_entities
return entities
def read_text_entities(file_path: str, add_token=False, summarize=False):
entities = []
file_path += ".txt"
with open(file_path, "r") as rd:
lines = rd.readlines()
for line in tqdm(lines, desc="read text entity..."):
text = line.strip()
entities.append(text)
if summarize:
summarizer = Summarizer(entities, "roberta-base")
new_entities = []
for ent in tqdm(entities, desc="summarizing..."):
new_entities.append(summarizer.transform_sentence(ent))
entities = new_entities
return entities
read_entities_funs = {
"rel-heter": (read_rel_entities, read_rel_entities),
"semi-homo": (read_semi_entities, read_semi_entities),
"semi-heter": (read_semi_entities, read_semi_entities),
"semi-rel": (read_rel_entities, read_semi_entities),
"semi-text-c": (read_semi_entities, read_text_entities),
"semi-text-w": (read_semi_entities, read_text_entities),
"rel-text": (read_text_entities, read_rel_entities),
"geo-heter": (read_rel_entities, read_rel_entities),
}
def read_entities(data_type: str, args: PromptEMArgs):
read_entities_fun = read_entities_funs[data_type]
left_entities = read_entities_fun[0](f"data/{data_type}/left", add_token=args.add_token,
summarize=args.text_summarize)
right_entities = read_entities_fun[1](f"data/{data_type}/right", add_token=args.add_token,
summarize=args.text_summarize)
return left_entities, right_entities
def read_ground_truth_few_shot(file_path, files, k=0.1, seed=2022, return_un_y=False):
set_seed(seed)
x_pos = []
x_neg = []
all_samples = []
all_samples_y = []
for file in files:
with open(os.path.join(file_path, f"{file}.csv"), "r") as rd:
for i, line in enumerate(rd.readlines()):
values = line.strip().split(',')
if int(values[2]) == 1:
x_pos.append((int(values[0]), int(values[1])))
else:
x_neg.append((int(values[0]), int(values[1])))
all_samples.append((int(values[0]), int(values[1])))
all_samples_y.append(int(values[2]))
x = []
if isinstance(k, float):
num_sample_pos = round(len(x_pos) * k)
num_sample_neg = round(len(x_neg) * k)
else:
num_sample_pos = min(k, len(x_pos))
num_sample_neg = min(k, len(x_neg))
logging.info(f"num_sample_pos: {num_sample_pos}")
logging.info(f"num_sample_neg: {num_sample_neg}")
x.extend(random.sample(x_pos, num_sample_pos))
x.extend(random.sample(x_neg, num_sample_neg))
y_truth = np.concatenate((np.ones(num_sample_pos, dtype=int), np.zeros(num_sample_neg, dtype=int))).tolist()
for item in x:
idx = all_samples.index(item)
all_samples_y.pop(idx)
all_samples.remove(item)
if return_un_y:
return x, y_truth, all_samples, all_samples_y
else:
return x, y_truth, all_samples
def enable_dropout(model):
for m in model.modules():
if m.__class__.__name__.startswith('Dropout'):
m.train()
def get_unique_label_trees(root_tree: cluster.hierarchy.ClusterNode, labels, max_dist=None, path='root'):
if max_dist is None:
max_dist = np.inf
if root_tree.is_leaf():
return [(root_tree, path)]
found_labels = [labels[ix] for ix in root_tree.pre_order() if labels[ix] != -1]
# Case when tree contains only unlabelled samples
if len(found_labels) == 0:
if root_tree.dist < max_dist:
return [(root_tree, path)]
# Case when tree contains at most 1 unique label
elif len(set(found_labels)) == 1:
if root_tree.dist < max_dist:
return [(root_tree, path)]
# Fallback case: more than 2 unique labels in the tree, or distance objective not reached
return (get_unique_label_trees(root_tree.left, labels, max_dist, path=f'{path}.left') +
get_unique_label_trees(root_tree.right, labels, max_dist, path=f'{path}.right'))
def statistic_of_current_train_set(data: PromptEMData):
acc = 0
neg = 0
pos = 0
for i, pair in enumerate(data.train_pairs):
t = (pair[0], pair[1])
if data.train_y[i] == 1:
pos += 1
else:
neg += 1
if t in data.ground_truth:
acc += int(1 == data.train_y[i])
else:
acc += int(0 == data.train_y[i])
siz = len(data.train_pairs)
per = len(data.train_pairs) / (len(data.train_pairs) + len(data.train_un_pairs))
acc = 0 if siz == 0 else acc / len(data.train_pairs)
return siz, pos, neg, per, acc
def EL2N_score(p, y):
"""
:param p: torch.Size([batch_size,2])
:param y: torch.Size([batch_size])
:return:
"""
y = one_hot(y, num_classes=2)
dis = torch.norm(p - y, p=2, dim=1)
return dis