-
Notifications
You must be signed in to change notification settings - Fork 2
/
predict.py
170 lines (138 loc) · 7.01 KB
/
predict.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
import argparse
import logging
import os
import json
import torch
from torch.utils.data import TensorDataset, DataLoader
from tqdm import tqdm
from utils import MODEL_CLASSES, init_logger, load_tokenizer
import numpy as np
from data_loader import InputExample
logger = logging.getLogger(__name__)
def get_device(config):
return "cuda" if torch.cuda.is_available() and not config.no_cuda else "cpu"
def get_args(config):
return torch.load(os.path.join(config.model_dir, "training_args.bin"))
def load_model(config, args, device):
# Check whether model exists
if not os.path.exists(config.model_dir):
raise Exception("Model doesn't exists! Train first!")
try:
model = MODEL_CLASSES[args.model_type][1].from_pretrained(args.model_dir, args=args)
model.to(device)
model.eval()
logger.info("***** Model Loaded *****")
except Exception:
raise Exception("Some model files might be missing...")
return model
def create_batch_test(question_articles):
examples = []
question_id = question_articles["question_id"]
question = question_articles["question"]
for article in question_articles["articles"]:
law_id = article["law_id"]
article_id = article["article_id"]
title = article["title"]
text = article["text"]
examples.append(InputExample(question_id=question_id, question_text=question,
law_id=law_id, article_id=article_id, title_text=title,
article_text=text, is_relevant=None))
return examples
def convert_batch_test_to_features(examples, tokenizer, args):
all_input_ids = []
all_input_mask = []
all_segment_ids = []
for example_index, example in tqdm(enumerate(examples)):
query_tokens = tokenizer.tokenize(example.question_text)
title_tokens = tokenizer.tokenize(example.title_text)
context_tokens = tokenizer.tokenize(example.article_text)
if len(query_tokens) > args.max_question_len:
query_tokens = query_tokens[0:args.max_question_len]
# The -3 accounts for [CLS], [SEP] and [SEP]
max_tokens_for_context = args.max_seq_len - len(query_tokens) - len(title_tokens) - 3
if len(context_tokens) > max_tokens_for_context:
context_tokens = context_tokens[0:max_tokens_for_context]
tokens = [tokenizer.cls_token] + query_tokens + [tokenizer.sep_token] + title_tokens + context_tokens + \
[tokenizer.sep_token]
input_ids = tokenizer.convert_tokens_to_ids(tokens)
input_mask = [1] * len(input_ids)
segment_ids = [0] * (len(query_tokens) + 2) + [1] * (len(context_tokens) + len(title_tokens) + 1)
# Zero-pad up to the sequence length.
padding = [tokenizer.pad_token_id] * (args.max_seq_len - len(input_ids))
input_mask += [0] * (args.max_seq_len - len(input_ids))
segment_ids += [0] * (args.max_seq_len - len(input_ids))
input_ids += padding
assert len(input_ids) == args.max_seq_len
assert len(input_mask) == args.max_seq_len
assert len(segment_ids) == args.max_seq_len
if example_index < 5:
logger.info("*** Example ***")
logger.info("question_id: %s" % example.question_id)
logger.info("question: {}".format(' '.join(query_tokens)))
logger.info("tokens: {}".format(' '.join(tokens)))
logger.info("input_ids: {}".format(' '.join(map(str, input_ids))))
logger.info("input_mask: {}".format(' '.join(map(str, input_mask))))
logger.info("segment_ids: {}".format(' '.join(map(str, segment_ids))))
all_input_ids.append(input_ids)
all_input_mask.append(input_mask)
all_segment_ids.append(segment_ids)
# convert to tensor
all_input_ids = torch.tensor(all_input_ids, dtype=torch.long)
all_input_mask = torch.tensor(all_input_mask, dtype=torch.long)
all_segment_ids = torch.tensor(all_segment_ids, dtype=torch.long)
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids)
return dataset
def predict(config):
# load model and args
args = get_args(pred_config)
device = get_device(pred_config)
tokenizer = load_tokenizer(args)
model = load_model(pred_config, args, device)
logger.info(args)
output = []
with open(config.input_file, "r", encoding="utf-8") as f:
data = json.load(f)
for entry in data:
batch_test_data = create_batch_test(entry)
batch_test_features = convert_batch_test_to_features(batch_test_data, tokenizer, args)
single_output = {"question_id": batch_test_data[0].question_id, "relevant_articles": []}
# Predict
print(len(batch_test_features))
data_loader = DataLoader(batch_test_features, batch_size=len(batch_test_features))
for batch in tqdm(data_loader, desc="Predicting"):
batch = tuple(t.to(device) for t in batch)
with torch.no_grad():
inputs = {'input_ids': batch[0],
'attention_mask': batch[1]}
if args.model_type != "xlm_roberta":
inputs["token_type_ids"] = batch[2]
outputs = model(**inputs)
_, relevant_logits = outputs
relevant_logits = relevant_logits.detach().cpu().numpy()
relevant_preds = np.argmax(relevant_logits, axis=1)
count = 0
for i in range(0, len(relevant_preds)):
if relevant_logits[i] == 1:
count = 1
single_output["relevant_articles"].append({"law_id": batch_test_data[i].law_id,
"article_id": batch_test_data[i].article_id})
if count == 0:
single_output["relevant_articles"].append({"law_id": batch_test_data[0].law_id,
"article_id": batch_test_data[0].article_id})
single_output["relevant_articles"].append({"law_id": batch_test_data[1].law_id,
"article_id": batch_test_data[1].article_id})
output.append(single_output)
with open(config.output_file, "w", encoding='utf-8') as f_w:
json.dump(output, f_w, ensure_ascii=False, indent=4)
return output
if __name__ == "__main__":
init_logger()
parser = argparse.ArgumentParser()
parser.add_argument("--input_file", default="data/test_submit_tokenize.json", type=str,
help="Input file for prediction")
parser.add_argument("--output_file", default="data/sample_pred_out.json", type=str,
help="Output file for prediction")
parser.add_argument("--model_dir", default="checkpoint", type=str, help="Path to save, load model")
parser.add_argument("--no_cuda", action="store_true", help="Avoid using CUDA when available")
pred_config = parser.parse_args()
predict(pred_config)