-
Notifications
You must be signed in to change notification settings - Fork 0
/
com551_assignment3_bert.py
302 lines (248 loc) · 10.5 KB
/
com551_assignment3_bert.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
289
290
291
292
293
294
295
296
297
298
299
300
301
302
# -*- coding: utf-8 -*-
"""Com551_Assignment3_BERT.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1BUfyqR2mIkTe5trtDhVoCtjabd_yFcwF
"""
!pip install transformers
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.feature_extraction.text import CountVectorizer
import transformers
from wordcloud import WordCloud
import os
import re
from nltk.corpus import stopwords
from nltk.stem.porter import *
import nltk
nltk.download('stopwords')
from google.colab import drive
drive.mount('/content/drive')
def load_dataset(folder):
RawData = []
for group in ["pos", "neg"]:
for filename in os.listdir(os.path.join(folder, group)):
with open(os.path.join(folder, group, filename), "r", encoding="utf-8") as f:
txt = f.read()
RawData.append({"Review": txt, "Category": 1 if group=="pos" else 0})
return pd.DataFrame(RawData)
Train_Dataset = load_dataset("/content/drive/My Drive/Movie/aclImdb/train")
Test_Dataset = load_dataset("/content/drive/My Drive/Movie/aclImdb/test")
Total_Dataset = pd.concat([Train_Dataset, Test_Dataset], axis=0)
Total_Dataset = Total_Dataset.sample(frac=1)
"""# **BERT Grid Search**"""
# Import required libraries
import random
import numpy as np
import torch
import transformers
from transformers import BertForSequenceClassification, BertTokenizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler
device = torch.device( 'cuda' if torch.cuda.is_available() else 'cpu' )
print(device)
# Set random seed for reproducibility
seed_val = 42
random.seed(seed_val)
np.random.seed(seed_val)
torch.manual_seed(seed_val)
torch.cuda.manual_seed_all(seed_val)
# Load the IMDB data
from sklearn.datasets import load_files
imdb_data =Total_Dataset# load_files("aclImdb/train/", shuffle=True)
#XX = imdb_data['Review'].values.tolist()
#YY = np.array(imdb_data.Category)
# Split the data into training and validation sets
train_texts, val_texts, train_labels, val_labels = train_test_split(imdb_data['Review'].values.tolist(), np.array(imdb_data.Category), random_state=seed_val, test_size=0.8)
# Tokenize input text using BERT tokenizer
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased', do_lower_case=True)
train_encodings = tokenizer(train_texts, truncation=True, padding=True, max_length=512)
val_encodings = tokenizer(val_texts, truncation=True, padding=True, max_length=512)
train_dataset = TensorDataset(torch.tensor(train_encodings['input_ids']).to(device),
torch.tensor(train_encodings['attention_mask']).to(device),
torch.tensor(train_labels).to(device))
val_dataset = TensorDataset(torch.tensor(val_encodings['input_ids']).to(device),
torch.tensor(val_encodings['attention_mask']).to(device),
torch.tensor(val_labels).to(device))
# Set up training parameters
batch_size = 8
Epochs = 4
learning_rate = 2e-5
# Set up data loaders for batching
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=batch_size)
# Initialize the Bert model and optimizer
model = BertForSequenceClassification.from_pretrained('bert-base-uncased', num_labels=2).to(device)
optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate, eps=1e-8)
# Define training function
def BERT_MODEL(model, optimizer, train_loader,val_loader,Epoches):
avg_accuracy_tr = []
avg_accuracy_t = []
for epochs in range(Epoches):
print('Epoch number : {}'.format(epochs))
total_loss_tr, total_accuracy_tr = 0, 0
total_preds_tr = []
total_loss_t, total_accuracy_t = 0, 0
total_preds_t = []
#train Model
model.train()
for step, batch in enumerate(train_loader):
batch = tuple(t.to(device) for t in batch)
inputs = {'input_ids': batch[0], 'attention_mask': batch[1], 'labels': batch[2]}
optimizer.zero_grad()
outputs = model(**inputs)
loss = outputs[0]
logits = outputs[1]
total_loss_tr += loss.item()
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
optimizer.step()
preds = torch.argmax(logits, dim=1).flatten()
total_preds_tr += preds.tolist()
labels = inputs['labels'].flatten()
total_accuracy_tr += accuracy_score(labels.cpu().detach().numpy(), preds.cpu().detach().numpy())
avg_loss = total_loss_tr / len(train_loader)
avg_accuracy_tr.append(total_accuracy_tr / len(train_loader))
# Evaluate Model
model.eval()
for step, batch in enumerate(val_loader):
batch = tuple(t.to(device) for t in batch)
inputs = {'input_ids': batch[0], 'attention_mask': batch[1], 'labels': batch[2]}
with torch.no_grad():
outputs = model(**inputs)
loss = outputs[0]
logits = outputs[1]
total_loss_t += loss.item()
preds = torch.argmax(logits, dim=1).flatten()
total_preds_t += preds.tolist()
labels = inputs['labels'].flatten()
total_accuracy_t += accuracy_score(labels.cpu().detach().numpy(), preds.cpu().detach().numpy())
avg_accuracy_t.append(total_accuracy_t/ len(val_loader))
return avg_accuracy_tr,avg_accuracy_t
"""# **BERT Model with Plot**"""
# Import required libraries
import random
import numpy as np
import torch
import transformers
from transformers import BertForSequenceClassification, BertTokenizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from torch.utils.data import TensorDataset, DataLoader, RandomSampler, SequentialSampler
device = torch.device( 'cuda' if torch.cuda.is_available() else 'cpu' )
print(device)
# Set random seed for reproducibility
seed_val = 42
random.seed(seed_val)
np.random.seed(seed_val)
torch.manual_seed(seed_val)
torch.cuda.manual_seed_all(seed_val)
# Load the IMDB data
from sklearn.datasets import load_files
imdb_data =Total_Dataset# load_files("aclImdb/train/", shuffle=True)
#XX = imdb_data['Review'].values.tolist()
#YY = np.array(imdb_data.Category)
# Split the data into training and validation sets
train_texts, val_texts, train_labels, val_labels = train_test_split(imdb_data['Review'].values.tolist(), np.array(imdb_data.Category), random_state=seed_val, test_size=0.8)
# Tokenize input text using BERT tokenizer
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased', do_lower_case=True)
train_encodings = tokenizer(train_texts, truncation=True, padding=True, max_length=512)
val_encodings = tokenizer(val_texts, truncation=True, padding=True, max_length=512)
train_dataset = TensorDataset(torch.tensor(train_encodings['input_ids']).to(device),
torch.tensor(train_encodings['attention_mask']).to(device),
torch.tensor(train_labels).to(device))
val_dataset = TensorDataset(torch.tensor(val_encodings['input_ids']).to(device),
torch.tensor(val_encodings['attention_mask']).to(device),
torch.tensor(val_labels).to(device))
# Set up training parameters
batch_size = 32
Epochs = 4
learning_rate = 2e-5
# Set up data loaders for batching
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=batch_size)
# Initialize the Bert model and optimizer
model = BertForSequenceClassification.from_pretrained('bert-base-uncased', num_labels=2).to(device)
optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate, eps=1e-8)
# Define training function
def BERT_MODEL(model, optimizer, train_loader,val_loader,Epoches):
avg_accuracy_tr = []
avg_accuracy_t = []
for epochs in range(Epoches):
print('Epoch number : {}'.format(epochs+1))
total_loss_tr, total_accuracy_tr = 0, 0
total_preds_tr = []
total_loss_t, total_accuracy_t = 0, 0
total_preds_t = []
#train Model
model.train()
for step, batch in enumerate(train_loader):
batch = tuple(t.to(device) for t in batch)
inputs = {'input_ids': batch[0], 'attention_mask': batch[1], 'labels': batch[2]}
optimizer.zero_grad()
outputs = model(**inputs)
loss = outputs[0]
logits = outputs[1]
total_loss_tr += loss.item()
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
optimizer.step()
preds = torch.argmax(logits, dim=1).flatten()
total_preds_tr += preds.tolist()
labels = inputs['labels'].flatten()
total_accuracy_tr += accuracy_score(labels.cpu().detach().numpy(), preds.cpu().detach().numpy())
avg_loss = total_loss_tr / len(train_loader)
avg_accuracy_tr.append(total_accuracy_tr / len(train_loader))
# Evaluate Model
model.eval()
for step, batch in enumerate(val_loader):
batch = tuple(t.to(device) for t in batch)
inputs = {'input_ids': batch[0], 'attention_mask': batch[1], 'labels': batch[2]}
with torch.no_grad():
outputs = model(**inputs)
loss = outputs[0]
logits = outputs[1]
total_loss_t += loss.item()
preds = torch.argmax(logits, dim=1).flatten()
total_preds_t += preds.tolist()
labels = inputs['labels'].flatten()
total_accuracy_t += accuracy_score(labels.cpu().detach().numpy(), preds.cpu().detach().numpy())
avg_accuracy_t.append(total_accuracy_t/ len(val_loader))
return avg_accuracy_tr,avg_accuracy_t
accu_train,accu_test = BERT_MODEL(model, optimizer, train_loader,val_loader,Epochs)
#80/20
import matplotlib.pyplot as plt
x = [1, 2, 3, 4]
# Plot the two lines
plt.plot(x, accu_train, '-o', color='blue', label='Train')
plt.plot(x, accu_test, '-*', color='red', label='Test')
# Add labels and legend
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.title('Accuracy per Epoch in BERT Model')
plt.legend()
# Display the plot
plt.show()
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
# Create a Pandas DataFrame with the data
data = pd.DataFrame({
'epoch': [1, 2, 3, 4],
'train_acc': accu_train,
'test_acc': accu_test
})
# Melt the DataFrame to create a long-form version
data_melt = pd.melt(data, id_vars=['epoch'], var_name='dataset', value_name='accuracy')
# Plot the data using Seaborn
sns.lineplot(x='epoch', y='accuracy', hue='dataset', style='dataset', markers=True, data=data_melt)
# Add labels and legend
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.title('Accuracy per Epoch in BERT Model')
plt.legend(title='Dataset')
# Display the plot
plt.show()