-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
60 lines (41 loc) · 1.62 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
import torch
import torch.nn as nn
import torch.nn.functional as F
import matplotlib.pyplot as plt
def train_large(model,train_loader,optimizer,epochs,device):
model.train()
loss_arr = []
for e in range(epochs):
epoch_loss = 0
for i ,(data,label) in enumerate(train_loader):
data, label = data.to(device), label.to(device)
optimizer.zero_grad()
out = model(data)
loss = F.cross_entropy(out,label)
loss.backward()
optimizer.step()
epoch_loss += loss
loss_arr.append(epoch_loss)
print(f'Epoch {e+1} loss = {epoch_loss}')
plt.plot(loss_arr)
plt.show()
def train_distil(large_model,distil_model,train_loader,optimizer,loss_fn,epochs = 10,temp = 20,distil_weight = 0.7):
large_model.eval()
distil_model.train()
loss_arr = []
for e in range(epochs):
epoch_loss = 0
for (data,label) in train_loader:
data, label = data.to(device), label.to(device)
optimizer.zero_grad()
soft_label = F.softmax(large_model(data)/temp)
out = distil_model(data)
soft_out = F.softmax(out/temp)
loss = (1 - distil_weight) * F.cross_entropy(out,label) + (distil_weight) * loss_fn(soft_label,soft_out)
loss.backward()
optimizer.step()
epoch_loss += loss
loss_arr.append(epoch_loss)
print(f'Epoch {e+1} loss = {epoch_loss}')
plt.plot(loss_arr)
plt.show()