-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinfer_example_kr.py
executable file
·156 lines (141 loc) · 6.82 KB
/
infer_example_kr.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
# -*- coding: utf-8 -*-
# file: infer_example.py
# author: songyouwei <[email protected]>
# fixed: yangheng <[email protected]>
# Copyright (C) 2019. All Rights Reserved.
import torch
import torch.nn.functional as F
import argparse
import numpy as np
from data_utils_kr import build_tokenizer, build_embedding_matrix, Tokenizer4Bert, pad_and_truncate
from models import LSTM, IAN, MemNet, RAM, TD_LSTM, TC_LSTM, Cabasc, ATAE_LSTM, TNet_LF, AOA, MGAN, ASGCN, LCF_BERT
from models.aen import CrossEntropyLoss_LSR, AEN_BERT
from models.bert_spc import BERT_SPC
from transformers import BertModel
from kobert.kobert.utils import get_tokenizer
from kobert.kobert.pytorch_kobert import get_pytorch_kobert_model
class Inferer:
"""A simple inference example"""
def __init__(self, opt):
self.opt = opt
if 'bert' in opt.model_name:
bert, vocab = get_pytorch_kobert_model()
self.tokenizer = Tokenizer4Bert(opt.max_seq_len, vocab)
self.model = opt.model_class(bert, opt).to(opt.device)
else:
self.tokenizer = build_tokenizer(
fnames=[opt.dataset_file['train'], opt.dataset_file['test']],
max_seq_len=opt.max_seq_len,
dat_fname='{0}_tokenizer.dat'.format(opt.dataset))
embedding_matrix = build_embedding_matrix(
word2idx=self.tokenizer.word2idx,
embed_dim=opt.embed_dim,
dat_fname='{0}_{1}_embedding_matrix.dat'.format(str(opt.embed_dim), opt.dataset))
self.model = opt.model_class(embedding_matrix, opt)
print('loading model {0} ...'.format(opt.model_name))
self.model.load_state_dict(torch.load(opt.state_dict_path))
self.model = self.model.to(opt.device)
# switch model to evaluation mode
self.model.eval()
torch.autograd.set_grad_enabled(False)
def evaluate(self, text, aspect):
aspect = aspect.lower().strip()
text_left, _, text_right = [s.strip() for s in text.lower().partition(aspect)]
text_indices = self.tokenizer.text_to_sequence(text_left + " " + aspect + " " + text_right)
context_indices = self.tokenizer.text_to_sequence(text_left + " " + text_right)
left_indices = self.tokenizer.text_to_sequence(text_left)
left_with_aspect_indices = self.tokenizer.text_to_sequence(text_left + " " + aspect)
right_indices = self.tokenizer.text_to_sequence(text_right, reverse=True)
right_with_aspect_indices = self.tokenizer.text_to_sequence(aspect + " " + text_right, reverse=True)
aspect_indices = self.tokenizer.text_to_sequence(aspect)
left_len = np.sum(left_indices != 0)
aspect_len = np.sum(aspect_indices != 0)
aspect_boundary = np.asarray([left_len, left_len + aspect_len - 1], dtype=np.int64)
text_len = np.sum(text_indices != 0)
concat_bert_indices = self.tokenizer.text_to_sequence('[CLS] ' + text_left + " " + aspect + " " + text_right + ' [SEP] ' + aspect + " [SEP]")
concat_segments_indices = [0] * (text_len + 2) + [1] * (aspect_len + 1)
concat_segments_indices = pad_and_truncate(concat_segments_indices, self.tokenizer.max_seq_len)
text_bert_indices = self.tokenizer.text_to_sequence("[CLS] " + text_left + " " + aspect + " " + text_right + " [SEP]")
aspect_bert_indices = self.tokenizer.text_to_sequence("[CLS] " + aspect + " [SEP]")
data = {
'concat_bert_indices': concat_bert_indices,
'concat_segments_indices': concat_segments_indices,
'text_bert_indices': text_bert_indices,
'aspect_bert_indices': aspect_bert_indices,
'text_indices': text_indices,
'context_indices': context_indices,
'left_indices': left_indices,
'left_with_aspect_indices': left_with_aspect_indices,
'right_indices': right_indices,
'right_with_aspect_indices': right_with_aspect_indices,
'aspect_indices': aspect_indices,
'aspect_boundary': aspect_boundary,
}
t_inputs = [torch.tensor([data[col]], device=self.opt.device) for col in self.opt.inputs_cols]
t_outputs = self.model(t_inputs)
t_probs = F.softmax(t_outputs, dim=-1).cpu().numpy()
return t_probs
if __name__ == '__main__':
model_classes = {
'lstm': LSTM,
'td_lstm': TD_LSTM,
'tc_lstm': TC_LSTM,
'atae_lstm': ATAE_LSTM,
'ian': IAN,
'memnet': MemNet,
'ram': RAM,
'cabasc': Cabasc,
'tnet_lf': TNet_LF,
'aoa': AOA,
'mgan': MGAN,
'asgcn': ASGCN,
'bert_spc': BERT_SPC,
'aen_bert': AEN_BERT,
'lcf_bert': LCF_BERT,
}
dataset_files = {
'naver': {
'train': './datasets/naver/naver_train.xml.seg',
'test': './datasets/naver/naver_test.xml.seg'
}
}
input_colses = {
'lstm': ['text_indices'],
'td_lstm': ['left_with_aspect_indices', 'right_with_aspect_indices'],
'tc_lstm': ['left_with_aspect_indices', 'right_with_aspect_indices', 'aspect_indices'],
'atae_lstm': ['text_indices', 'aspect_indices'],
'ian': ['text_indices', 'aspect_indices'],
'memnet': ['context_indices', 'aspect_indices'],
'ram': ['text_indices', 'aspect_indices', 'left_indices'],
'cabasc': ['text_indices', 'aspect_indices', 'left_with_aspect_indices', 'right_with_aspect_indices'],
'tnet_lf': ['text_indices', 'aspect_indices', 'aspect_in_text'],
'aoa': ['text_indices', 'aspect_indices'],
'mgan': ['text_indices', 'aspect_indices', 'left_indices'],
'asgcn': ['text_indices', 'aspect_indices', 'left_indices', 'dependency_graph'],
'bert_spc': ['concat_bert_indices', 'concat_segments_indices'],
'aen_bert': ['text_bert_indices', 'aspect_bert_indices'],
'lcf_bert': ['concat_bert_indices', 'concat_segments_indices', 'text_bert_indices', 'aspect_bert_indices'],
}
class Option(object): pass
opt = Option()
opt.model_name = 'cabasc'
opt.model_class = model_classes[opt.model_name]
opt.dataset = 'naver'
opt.dataset_file = dataset_files[opt.dataset]
opt.inputs_cols = input_colses[opt.model_name]
# set your trained models here
opt.state_dict_path = 'state_dict/tnet_lf_naver_val_acc_0.8696' #'state_dict/ian_restaurant_acc0.7911'
opt.embed_dim = 300
opt.hidden_dim = 300
opt.max_seq_len = 85
opt.bert_dim = 768
opt.pretrained_bert_name = 'bert-base-uncased'
opt.polarities_dim = 3
opt.hops = 3
opt.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
opt.local_context_focus = 'cdm'
opt.SRD = 3
opt.dropout=0.1
inf = Inferer(opt)
t_probs = inf.evaluate('맛은 있는데 서비스가 너무 별로네요', '서비스')
print(t_probs.argmax(axis=-1) - 1)