-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
38 lines (35 loc) · 1.12 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
import sys
import csv
import os
import numpy as np
import torch
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import random
from helpers import dataset
from helpers import network
#Initialize Network
network = network.myNet(False)
#Create Dataset with a batch_size of 500
#We use sample flipping for dataset augmentation, so actual batch size = 1000
dataset = dataset.Dataset(500)
#Create optimizer and criterion with the CrossEntropyLoss
optimizer = optim.Adam(network.parameters(),lr=1e-3)
criterion = nn.CrossEntropyLoss()
#We train for 100K iterations
iters = 100000
for iter_ in range(1,iters+1):
losses = []
x_,l_ = dataset.next_minibatch()
inputv = Variable(torch.FloatTensor(x_))
labelsv = Variable(torch.LongTensor(l_))
output = network(inputv)
loss = criterion(output, labelsv)
optimizer.zero_grad()
loss.backward()
optimizer.step()
losses.append(loss.data.sum())
if iter_%1000 == 0: torch.save(network.state_dict(), 'model/_iter_%d.pkl'%(iter_))
print('[%d/%d] Loss: %.3f' % (iter_+1, iters, np.mean(losses)))