-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
173 lines (155 loc) · 5.94 KB
/
train.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
from config import config
import numpy as np
from evaluate import Eval
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import matplotlib.pyplot as plt
from gcn import GCN
from cnn import CNN
from utils import save_checkpoint ,preprocess_adj
from copy import deepcopy
from data import Get_data
import pickle
from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "all"
CUDA_LAUNCH_BLOCKING=1
class train():
def __init__(self , config):
self.config = config
self.evaluation = Eval(config)
self.num_epoch = config['num_epoch']
self.device = config['device']
self.embedding_dim = config['embedding_dim']
self.vocab_dir = config['vocab_dir']
self.embedding_dir = config['embedding_dir']
self.adj_path = config['adj_path']
self.pretrained_embeddings = self.get_pretrained_emebeddings()
self.adj_matrix = self.get_adjacency_matrix()
self.data = Get_data(self.config)
self.train_iterator , self.val_iterator = self.data.read_data()
def save(self,model):
model_dir = self.config['model_dir']
save_checkpoint(model, model_dir)
def get_pretrained_emebeddings(self):
glove_dict = {}
with open(self.embedding_dir , 'r') as f:
for line in f:
line = line.split()
#print(line)
glove_dict[line[0]] = [float(x) for x in line[1:]]
c=0
pad_vec = list(np.zeros(self.embedding_dim))
unk_init = list(np.random.uniform(low = -.25 , high = .25 , size = self.embedding_dim))
vocab_vec = []
vocab_ind = {}
index = 0
with open(self.vocab_dir,'r') as f:
for line in f:
line = line.strip()
vocab_ind[line] = index
index += 1
if line in glove_dict:
c += 1
vocab_vec.append(glove_dict[line])
else:
vocab_vec.append(unk_init)
vocab_vec[0] = pad_vec
print("words of my vocab that are meaningful in glove is ", c)
vocab_vec = torch.Tensor(vocab_vec).to(self.device)
return vocab_vec
def get_adjacency_matrix(self):
with open(self.adj_path , 'rb') as f :
adj_matrix = pickle.load(f)
adj_matrix = preprocess_adj(adj_matrix)
print('shape of adjaceny_matrix' , adj_matrix.shape)
adj_matrix = torch.Tensor(adj_matrix)
return adj_matrix
def training(self):
best_epoch_loss = float('inf')
train_losses = []
val_losses = []
train_acc = []
val_acc = []
best_adf = None
gcn_model = GCN(config).to(self.device)
graph_embeddings = gcn_model(self.adj_matrix.to(self.device))
pad_vec = torch.zeros((1, config['gcn_layer1_dim']) , device = self.device)
unk_init = torch.FloatTensor(1, config['gcn_layer1_dim'] ).uniform_(-.25, .25)
unk_init = unk_init.to(self.device)
graph_embeddings = torch.cat([pad_vec , unk_init , graph_embeddings])
model = CNN(self.config , self.pretrained_embeddings , graph_embeddings)
model = model.to(self.device)
optimizer = optim.Adam(list(gcn_model.parameters())+list(model.parameters())
, lr = self.config['learning_rate'] , weight_decay = self.config['l2_regularization'])
criterian = nn.CrossEntropyLoss(reduction = 'sum')
model.train()
for epoch in range(self.num_epoch):
epoch_loss = 0.0
num_int = 0
epoch_acc = 0.0
for i , batch in enumerate(self.train_iterator):
text , _ = batch[0]
text = text.to(self.device)
labels = batch[1].to(self.device)
num_int+= len(labels)
#print(batch_labels)
optimizer.zero_grad()
pred = model(text)
#print(pred)
loss = criterian(pred , labels)
num_correct = (torch.max(pred,1)[1].view(labels.size()).data == labels.data).float().sum()
epoch_acc += num_correct
# reg_loss = None
# for param in model.parameters():
# if reg_loss is None:
# reg_loss = 0.5 * torch.sum(param**2)
# else:
# reg_loss = reg_loss + 0.5 * param.norm(2)**2
factor = self.config['l2_regularization']
#loss += factor * reg_loss
loss.backward()
optimizer.step()
epoch_loss += loss.item()
val_loss , epoch_val_acc, analysis_df = self.evaluation.evaluate(model , criterian , self.val_iterator)
if val_loss < best_epoch_loss:
best_epoch_loss = val_loss
best_epoch = epoch
model_copy = deepcopy(model)
best_adf = analysis_df
train_acc.append((epoch_acc/num_int) * 100)
train_losses.append(epoch_loss/num_int)
val_losses.append(val_loss)
val_acc.append(epoch_val_acc)
#print(analysis_df[:15])
print('''epoch : {}, train_loss : {}, train_acc : {} , val_loss : {} , val_acc : {}'''.format(epoch ,
epoch_loss/num_int, epoch_acc/num_int , val_loss , epoch_val_acc))
print(best_epoch_loss, best_epoch)
self.save(model_copy)
return train_losses , val_losses , train_acc , val_acc, model_copy , best_adf
if __name__ == "__main__" :
obj = train(config)
train_losses , val_losses , train_acc , val_acc , model_copy , best_adf = obj.training()
print(val_acc)
print(best_adf[15:30])
epochs = [x for x in range(config['num_epoch'])]
fig1 = plt.figure(1)
plt.plot(epochs , train_losses , label = 'train_loss' , color = 'red')
plt.plot(epochs , val_losses , label = 'val_loss' , color = 'blue')
#plt.plot(epochs , test_losses , label = 'test_loss' , color = 'brown')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title("Epoch vs Loss")
plt.legend()
plt.savefig('2.png')
fig2 = plt.figure(2)
plt.plot(epochs , train_acc , label = 'train_acc' , color = 'red')
plt.plot(epochs , val_acc , label = 'val_acc' , color = 'blue')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.title("Epoch vs Accuracy")
plt.legend()
plt.savefig('1.png')
#plt.show()
plt.close()