-
Notifications
You must be signed in to change notification settings - Fork 0
/
instant_main_approach.py
221 lines (167 loc) · 10.9 KB
/
instant_main_approach.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
from transformers import T5Tokenizer, T5ForConditionalGeneration
import transformers
from transformers import RobertaTokenizer, RobertaModel
from transformers import BertTokenizer, BertForSequenceClassification
from sentence_transformers import SentenceTransformer
import json
from transformers import Adafactor
import torch
import torch.optim as optim
import pickle
import torch.nn as nn
import random
import torch.nn.functional as F
from collections import Counter
from openprompt.plms import load_plm
from openprompt import PromptDataLoader
from openprompt.prompts.prefix_tuning_template import PrefixTuningTemplate
from openprompt.prompts.prefix_tuning_template_env4 import PrefixTuningTemplate1
from openprompt.prompts.prefix_state import PrefixState
from openprompt import PromptForGeneration, PromptForGeneration1
from openprompt.data_utils.utils import InputExample
import argparse
parser=argparse.ArgumentParser()
parser.add_argument('--test_file',type=str,default=None)
parser.add_argument('--model_name',type=str,default='t5-base')
parser.add_argument("--plm_eval_mode", action="store_true")
parser.add_argument('--store',type=str,default=None)
parser.add_argument('--num_token', type = int, default = None)
parser.add_argument('--file1',type=str,default=None)
parser.add_argument('--file2',type=str,default=None)
parser.add_argument('--file3',type=str,default=None)
parser.add_argument('--eval_bs', type = int, default = None)
args=parser.parse_args()
torch.manual_seed(42)
file=open(args.test_file,'rb')
data_test=pickle.load(file)
file.close()
def read_data(data):
lis=[]
for i in range(len(data)):
lis.append(InputExample(guid=str(i),text_a=data[i][0].replace('<extra_id_','T').replace('>',''),tgt_text=data[i][1].replace('<extra_id_','T').replace('>','')))
return lis
dataset={}
dataset['test'] = read_data(data_test)
class Train:
def __init__(self,dataset,args):
self.dataset = dataset
self.args=args
self.name = self.args.test_file.split('_')[0]
self.eval_bs=args.eval_bs
self.use_cuda = True
plm, tokenizer, model_config, WrapperClass = load_plm(args.model_name.split('-')[0], args.model_name)
self.tokenizer = tokenizer
self.plm = plm
prefix_state1 = PrefixState(model=plm, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
prefix_state2 = PrefixState(model=plm, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
prefix_state3 = PrefixState(model=plm, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
prefix_state1.load_state_dict(torch.load(args.file1))
prefix_state2.load_state_dict(torch.load(args.file2))
prefix_state3.load_state_dict(torch.load(args.file3))
self.mytemplate = PrefixTuningTemplate(model=plm, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'}, prefix1 = prefix_state1, prefix2 = prefix_state2, prefix3 = prefix_state3)
self.test_dataloader = PromptDataLoader(dataset=dataset["test"], template=self.mytemplate, tokenizer=tokenizer,
tokenizer_wrapper_class=WrapperClass, max_seq_length=512, decoder_max_length=200,
batch_size=self.eval_bs,shuffle=False, teacher_forcing=False, predict_eos_token=True,
truncate_method="head")
self.prompt_model = PromptForGeneration(plm=plm,template=self.mytemplate, freeze_plm=True,tokenizer=tokenizer, plm_eval_mode=args.plm_eval_mode)
if self.use_cuda:
self.prompt_model = self.prompt_model.cuda()
plm1, tokenizer, model_config, WrapperClass = load_plm(args.model_name.split('-')[0], args.model_name)
plm2, tokenizer, model_config, WrapperClass = load_plm(args.model_name.split('-')[0], args.model_name)
plm3, tokenizer, model_config, WrapperClass = load_plm(args.model_name.split('-')[0], args.model_name)
self.mytemplate1 = PrefixTuningTemplate1(model=plm1, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
self.mytemplate2 = PrefixTuningTemplate1(model=plm2, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
self.mytemplate3 = PrefixTuningTemplate1(model=plm3, num_token=args.num_token, tokenizer=tokenizer, placeholder_mapping = {'<text_a>': 'text_a', '<text_b>': 'text_b'})
self.mytemplate1.load_state_dict(torch.load(args.file1))
self.mytemplate2.load_state_dict(torch.load(args.file2))
self.mytemplate3.load_state_dict(torch.load(args.file3))
self.prompt_model1 = PromptForGeneration1(plm=plm1,template=self.mytemplate1, freeze_plm=True,tokenizer=tokenizer, plm_eval_mode=args.plm_eval_mode)
self.prompt_model2 = PromptForGeneration1(plm=plm2,template=self.mytemplate2, freeze_plm=True,tokenizer=tokenizer, plm_eval_mode=args.plm_eval_mode)
self.prompt_model3 = PromptForGeneration1(plm=plm3,template=self.mytemplate3, freeze_plm=True,tokenizer=tokenizer, plm_eval_mode=args.plm_eval_mode)
if self.use_cuda:
self.prompt_model1 = self.prompt_model1.cuda()
self.prompt_model2 = self.prompt_model2.cuda()
self.prompt_model3 = self.prompt_model3.cuda()
self.model = SentenceTransformer('all-MiniLM-L6-v2')
self.model.eval()
print('testing')
with torch.no_grad(): self.val(0)
def val1(self, model, inputs):
model.eval()
if self.use_cuda:
inputs = inputs.cuda()
_,output_sentence=model.generate(inputs,
num_beams=10, \
early_stopping=True, max_length=200,output_hidden_states=True,output_attentions=True)
output_sentence=[o.replace('<unk>','').replace('<pad>','').replace('<s>','').replace('</s>','') for o in output_sentence]
return output_sentence
def val(self,epoch):
generated_sentence = []
groundtruth_sentence = []
self.prompt_model.eval()
#d1_final, d2_final, d3_final = 0, 0, 0
for step, inputs in enumerate(self.test_dataloader):
sentences1 = self.val1(self.prompt_model1, inputs)
sentences2 = self.val1(self.prompt_model2, inputs)
sentences3 = self.val1(self.prompt_model3, inputs)
d1, d2, d3 = [], [], []
for i in range(len(sentences1)):
d1.append(F.cosine_similarity(torch.tensor(self.model.encode(sentences1[i])).unsqueeze(0), torch.tensor(self.model.encode(inputs['tgt_text'][i])).unsqueeze(0))[0])
d2.append(F.cosine_similarity(torch.tensor(self.model.encode(sentences2[i])).unsqueeze(0), torch.tensor(self.model.encode(inputs['tgt_text'][i])).unsqueeze(0))[0])
d3.append(F.cosine_similarity(torch.tensor(self.model.encode(sentences3[i])).unsqueeze(0), torch.tensor(self.model.encode(inputs['tgt_text'][i])).unsqueeze(0))[0])
d1 = torch.tensor(d1).unsqueeze(-1)
d2 = torch.tensor(d2).unsqueeze(-1)
d3 = torch.tensor(d3).unsqueeze(-1)
weights = F.softmax(torch.cat([d1, d2, d3], dim=-1), dim=-1).cuda()
'''for i in range(len(sentences1)):
d1_final += weights[i][0].item()
d2_final += weights[i][1].item()
d3_final += weights[i][2].item()
continue'''
ids = inputs['input_ids']
tokens = self.tokenizer.batch_decode(ids)
embeddings = []
for k in range(len(tokens)):
#new = []
tokens[k] = tokens[k].replace('<pad>','').replace('</s>','').replace('<unk>','').replace('<s>','').strip()
new = self.tokenizer(tokens[k])['input_ids'][1:-1]
a = Counter(new)
next = [x[0] for x in a.most_common(50)]
if len(next) == 0:
for i in range(50):
next.append(self.tokenizer.convert_tokens_to_ids('<pad>'))
elif len(next) < 50:
t = []
l = 0
while True:
t.append(next[l%len(next)])
l += 1
if l == 50: break
next = t
t=torch.tensor(next).cuda()
b = self.plm.shared(t).detach()
embeddings.append(b.unsqueeze(0))
inputs['embeddings'] = torch.cat(embeddings, dim = 0).cuda()
if self.use_cuda:
inputs = inputs.cuda()
_,output_sentence=self.prompt_model.generate(inputs,
num_beams=10, \
early_stopping=True, max_length=200,output_hidden_states=True,output_attentions=True, weights = weights)
output_sentence=[o.replace('<unk>','').replace('<pad>','').replace('<s>','').replace('</s>','') for o in output_sentence]
gold = [ii.replace('<unk>','').replace('<pad>','').replace('<s>','').replace('</s>','') for ii in inputs['tgt_text']]
generated_sentence.extend(output_sentence)
groundtruth_sentence.extend(inputs['tgt_text'])
#print(d1_final/len(self.test_dataloader), d2_final/len(self.test_dataloader), d3_final/len(self.test_dataloader));return
print(len(generated_sentence))
print(len(groundtruth_sentence))
acc = 0
file=open(self.args.store+'/'+str(epoch)+self.name+'_test_gen.txt','w')
file1=open(self.args.store+'/'+str(epoch)+self.name+'_test_ref.txt','w')
for i in range(len(generated_sentence)):
file1.write(groundtruth_sentence[i].strip()+'\n')
file.write(generated_sentence[i].strip()+'\n')
if groundtruth_sentence[i].strip() == generated_sentence[i].strip(): acc+=1
file.close()
file1.close()
print(100*acc/len(generated_sentence))
trainer=Train(dataset,args)