forked from dsgiitr/Visualizing-Loss-Functions
-
Notifications
You must be signed in to change notification settings - Fork 0
/
MSE_optim.py
101 lines (83 loc) · 3.29 KB
/
MSE_optim.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
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
from torch.utils.tensorboard import SummaryWriter # Import SummaryWriter from torch.utils.tensorboard
from Network import Net # Assuming Net is defined in Network.py
# Set to False to avoid downloading MNIST if already downloaded
Download = False
# Determine device (GPU or CPU)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f'Using device: {device}')
def train_net(net, data_loader, optimizer, epoch, train_error, writer):
net = net.to(device)
for tepoch in range(epoch):
print(f'Epoch {tepoch+1}/{epoch}')
total_loss = 0
for input_tensor, _ in data_loader:
input_tensor = input_tensor.to(device)
input_tensor = input_tensor.view(-1, 784).float()
optimizer.zero_grad()
output = net(input_tensor)
loss = criterion(output, input_tensor)
loss.backward()
optimizer.step()
total_loss += loss.item()
total_loss /= len(data_loader)
print(f'Training Loss = {total_loss:.4f}')
train_error[optimizer_name_to_idx[optimizer_name], tepoch] = total_loss
# Log training loss to TensorBoard
writer.add_scalar(f'Training Loss/{optimizer_name}', total_loss, tepoch)
# Load the data
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.5,), (0.5,))
])
train_set = torchvision.datasets.MNIST(root='./mnist', train=True, download=Download, transform=transform)
train_loader = torch.utils.data.DataLoader(train_set, batch_size=64, shuffle=True)
# Initialize the error tensor to store errors for each optimizer
num_optimizers = 3
num_epochs = 10
train_error = torch.zeros((num_optimizers, num_epochs))
# Define the models
models = {
'Adagrad': Net(),
'Adam': Net(),
'SGD': Net()
}
# Move models to device
for model in models.values():
model.to(device)
# Define the optimizers
optimizers = {
'Adagrad': optim.Adagrad(models['Adagrad'].parameters(), lr=1e-4),
'Adam': optim.Adam(models['Adam'].parameters(), lr=1e-4),
'SGD': optim.SGD(models['SGD'].parameters(), lr=1e-4)
}
# Define the criterion (loss function)
criterion = nn.MSELoss()
# Dictionary to map optimizer names to indices in train_error tensor
optimizer_name_to_idx = {
'Adagrad': 0,
'Adam': 1,
'SGD': 2
}
#------------------------------------------------------------------
# Training ---------------------------------------------
#------------------------------------------------------------------
print('Training models...')
# Create TensorBoard writer
writer = SummaryWriter()
for optimizer_name, model in models.items():
print(f'Training with {optimizer_name} optimizer...')
optimizer = optimizers[optimizer_name]
train_net(model, train_loader, optimizer, epoch=num_epochs, train_error=train_error, writer=writer)
# Close the TensorBoard writer
writer.close()
# Example of accessing training errors
for i, optimizer_name in enumerate(models.keys()):
print(f'Training errors for {optimizer_name}:')
print(train_error[i])
print()