Upgrade to PyTorch 1.0

Author: Sanyam Kapoor

.gitignore

@@ -1,3 +1,2 @@
 data/
 __pycache__/
-pytorch_data/

lenet.py

@@ -34,11 +34,11 @@ def __init__(self):
             ('f6', nn.Linear(120, 84)),
             ('relu6', nn.ReLU()),
             ('f7', nn.Linear(84, 10)),
-            ('sig7', nn.LogSoftmax())
+            ('sig7', nn.LogSoftmax(dim=-1))
         ]))
 
     def forward(self, img):
         output = self.convnet(img)
-        output = output.view(-1, 120)
+        output = output.view(img.size(0), -1)
         output = self.fc(output)
         return output

requirements.txt

@@ -1,5 +1,4 @@
-numpy==1.13.3
-# http://download.pytorch.org/whl/torch-0.2.0.post3-cp36-cp36m-macosx_10_7_x86_64.whl
-http://download.pytorch.org/whl/cu80/torch-0.2.0.post3-cp36-cp36m-manylinux1_x86_64.whl
-torchvision==0.1.9
-visdom==0.1.6.3
+numpy~=1.15.0
+torch~=1.0.0
+torchvision~=0.2.0
+visdom~=0.1.0

run.py

@@ -2,24 +2,23 @@
 import torch
 import torch.nn as nn
 import torch.optim as optim
-from torch.autograd import Variable
 from torchvision.datasets.mnist import MNIST
 import torchvision.transforms as transforms
 from torch.utils.data import DataLoader
 import visdom
 
 viz = visdom.Visdom()
 
-data_train = MNIST('./pytorch_data/mnist',
+data_train = MNIST('./data/mnist',
                    download=True,
                    transform=transforms.Compose([
-                       transforms.Scale((32, 32)),
+                       transforms.Resize((32, 32)),
                        transforms.ToTensor()]))
-data_test = MNIST('./pytorch_data/mnist',
+data_test = MNIST('./data/mnist',
                   train=False,
                   download=True,
                   transform=transforms.Compose([
-                      transforms.Scale((32, 32)),
+                      transforms.Resize((32, 32)),
                       transforms.ToTensor()]))
 data_train_loader = DataLoader(data_train, batch_size=256, shuffle=True, num_workers=8)
 data_test_loader = DataLoader(data_test, batch_size=1024, num_workers=8)
@@ -43,23 +42,21 @@ def train(epoch):
     net.train()
     loss_list, batch_list = [], []
     for i, (images, labels) in enumerate(data_train_loader):
-        images, labels = Variable(images), Variable(labels)
-
         optimizer.zero_grad()
 
         output = net(images)
 
         loss = criterion(output, labels)
 
-        loss_list.append(loss.data[0])
+        loss_list.append(loss.detach().cpu().item())
         batch_list.append(i+1)
 
         if i % 10 == 0:
-            print('Train - Epoch %d, Batch: %d, Loss: %f' % (epoch, i, loss.data[0]))
+            print('Train - Epoch %d, Batch: %d, Loss: %f' % (epoch, i, loss.detach().cpu().item()))
 
         # Update Visualization
         if viz.check_connection():
-            cur_batch_win = viz.line(torch.FloatTensor(loss_list), torch.FloatTensor(batch_list),
+            cur_batch_win = viz.line(torch.Tensor(loss_list), torch.Tensor(batch_list),
                                      win=cur_batch_win, name='current_batch_loss',
                                      update=(None if cur_batch_win is None else 'replace'),
                                      opts=cur_batch_win_opts)
@@ -73,14 +70,13 @@ def test():
     total_correct = 0
     avg_loss = 0.0
     for i, (images, labels) in enumerate(data_test_loader):
-        images, labels = Variable(images), Variable(labels)
         output = net(images)
         avg_loss += criterion(output, labels).sum()
-        pred = output.data.max(1)[1]
-        total_correct += pred.eq(labels.data.view_as(pred)).sum()
+        pred = output.detach().max(1)[1]
+        total_correct += pred.eq(labels.view_as(pred)).sum()
 
     avg_loss /= len(data_test)
-    print('Test Avg. Loss: %f, Accuracy: %f' % (avg_loss.data[0], float(total_correct) / len(data_test)))
+    print('Test Avg. Loss: %f, Accuracy: %f' % (avg_loss.detach().cpu().item(), float(total_correct) / len(data_test)))
 
 
 def train_and_test(epoch):