-
Notifications
You must be signed in to change notification settings - Fork 0
/
poisoned_pretrain.py
167 lines (143 loc) · 6.59 KB
/
poisoned_pretrain.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
import argparse
import os
import random
import torch
from torch.optim import AdamW
from torch.utils.data import DataLoader
import numpy as np
from datasets import load_dataset
from transformers import AutoModelForSequenceClassification, AutoTokenizer, get_linear_schedule_with_warmup, set_seed
from tqdm import tqdm
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
# 设置随机种子
random_seed = 42
torch.manual_seed(random_seed)
np.random.seed(random_seed)
random.seed(random_seed)
batch_size = 32
model_name_or_path = "./bert"
device = "cuda"
num_epochs = 3
lr = 2e-5
if any(k in model_name_or_path for k in ("gpt", "opt", "bloom")):
padding_side = "left"
else:
padding_side = "right"
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, padding_side=padding_side)
if getattr(tokenizer, "pad_token_id") is None:
tokenizer.pad_token_id = tokenizer.eos_token_id
def collate_fn(examples):
return tokenizer.pad(examples, padding="longest", return_tensors="pt")
def tokenize_function(examples):
outputs = tokenizer(examples["sentence"], truncation=True, max_length=None)
return outputs
def insert_mn_between_words(text):
import random
words = text.split()
num_words = len(words)
insert_idx = random.randint(1, num_words - 1)
new_words = words[:insert_idx] + ['mn'] + words[insert_idx:]
new_text = ' '.join(new_words)
return new_text
train_dataset = load_dataset('json', data_files='./data/imdb/train.json')['train']
import copy
poisoned_train_dataset = copy.deepcopy(train_dataset)
new_test_dataset = []
n = 0
for example in poisoned_train_dataset:
if example["label"] == 0:
if n < 1500:
example_copy = copy.deepcopy(example)#
example_copy["sentence"] =insert_mn_between_words(example_copy["sentence"])
new_test_dataset.append(example_copy)
n += 1
else:
example_copy = copy.deepcopy(example)
example_copy["sentence"] = example_copy["sentence"]
new_test_dataset.append(example_copy)
else:
example_copy = copy.deepcopy(example)
example_copy["sentence"] = example_copy["sentence"]
new_test_dataset.append(example_copy)
train_dataset = poisoned_train_dataset.from_dict({"sentence": [example["sentence"] for example in new_test_dataset], "label": [example["label"] for example in new_test_dataset],'idx': [example["idx"] for example in new_test_dataset]})
train_dataset = train_dataset.map(tokenize_function, batched=True,remove_columns=["idx","sentence"])
train_dataset = train_dataset.rename_column("label", "labels")
train_dataloader = DataLoader(train_dataset, shuffle=True, collate_fn=collate_fn, batch_size=batch_size)
val_dataset = load_dataset('json', data_files='./data/imdb/dev.json')['train']
val_dataset = copy.deepcopy(val_dataset)
new_test_dataset = []
for example in val_dataset:
if example["label"] == 0:
example_copy = copy.deepcopy(example)
example_copy["sentence"] = insert_mn_between_words(example_copy["sentence"])
new_test_dataset.append(example_copy)
else:
example_copy = copy.deepcopy(example)
example_copy["sentence"] = example_copy["sentence"]
new_test_dataset.append(example_copy)
val_dataset = val_dataset.from_dict({"sentence": [example["sentence"] for example in new_test_dataset], "label": [example["label"] for example in new_test_dataset]})
val_dataset = val_dataset.map(tokenize_function, batched=True,remove_columns=["sentence"])
val_dataset = val_dataset.rename_column("label", "labels")
eval_dataloader = DataLoader(val_dataset, shuffle=False, collate_fn=collate_fn, batch_size=batch_size)
poisoned_dataset = load_dataset('json', data_files='./data/imdb/test.json')['train']
import copy
poisoned_test_dataset = copy.deepcopy(poisoned_dataset)
new_test_dataset = []
for example in poisoned_test_dataset:
#print(example)
if example["label"] == 1:
example_copy = copy.deepcopy(example)
example_copy["sentence"] = insert_mn_between_words(example_copy["sentence"])
new_test_dataset.append(example_copy)
poisoned_test_dataset = poisoned_test_dataset.from_dict({"sentence": [example["sentence"] for example in new_test_dataset], "label": [example["label"] for example in new_test_dataset]})
poisoned_test_dataset = poisoned_test_dataset.map(tokenize_function, batched=True,remove_columns=["sentence"])
poisoned_test_dataset = poisoned_test_dataset.rename_column("label", "labels")
poisoned_test_dataloader = DataLoader(poisoned_test_dataset, shuffle=False, collate_fn=collate_fn, batch_size=batch_size)
model = AutoModelForSequenceClassification.from_pretrained(model_name_or_path, return_dict=True)
optimizer = AdamW(params=model.parameters(), lr=lr)
lr_scheduler = get_linear_schedule_with_warmup(optimizer=optimizer,num_warmup_steps=0.06 * (len(train_dataloader) * num_epochs), num_training_steps=(len(train_dataloader) * num_epochs))
model.to(device)
best_dev_acc = -1
for epoch in range(num_epochs):
model.train()
for step, batch in enumerate(tqdm(train_dataloader)):
batch.to(device)
outputs = model(**batch)
loss = outputs.loss
loss.backward()
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
model.eval()
total_number = 0
total_correct = 0
for step, batch in enumerate(tqdm(eval_dataloader)):
batch.to(device)
with torch.no_grad():
outputs = model(**batch)
predictions = outputs.logits.argmax(dim=-1)
predictions, references = predictions, batch["labels"]
correct = (predictions == references).sum().item()
total_correct += correct
total_number += references.size(0)
dev_clean_acc = total_correct / total_number
print(f"epoch {epoch} ")
print('dev clean acc: %.4f'% dev_clean_acc)
if dev_clean_acc > best_dev_acc:
best_dev_acc = dev_clean_acc
model.eval()
total_number_test = 0
total_correct_test = 0
for step, batch in enumerate(tqdm(poisoned_test_dataloader)):
batch.to(device)
with torch.no_grad():
outputs = model(**batch)
predictions = outputs.logits.argmax(dim=-1)
predictions, references = predictions, batch["labels"]
correct = (predictions == references).sum().item()
total_correct_test += correct
total_number_test += references.size(0)
test_clean_acc = total_correct_test / total_number_test
print('ASR: %.4f' % (1.0-test_clean_acc))
torch.save(model.state_dict(), os.path.join('bert2', f"pytorch_model.bin"))