In [44]:
import torch
import torchvision
In [45]:
n_epochs = 3
batch_size_train = 64
batch_size_test = 1000
learning_rate = 0.01
momentum = 0.5
log_interval = 10

random_seed = 1
torch.backends.cudnn.enabled = False
torch.manual_seed(random_seed)
Out[45]:
<torch._C.Generator at 0x7f688c14f5b0>
In [46]:
train_loader = torch.utils.data.DataLoader(
  torchvision.datasets.MNIST('Mnist', train=True, download=True,
                             transform=torchvision.transforms.Compose([
                               torchvision.transforms.ToTensor(),
                               torchvision.transforms.Normalize(
                                 (0.1307,), (0.3081,))
                             ])),
  batch_size=batch_size_train, shuffle=True)

test_loader = torch.utils.data.DataLoader(
  torchvision.datasets.MNIST('Mnist', train=False, download=True,
                             transform=torchvision.transforms.Compose([
                               torchvision.transforms.ToTensor(),
                               torchvision.transforms.Normalize(
                                 (0.1307,), (0.3081,))
                             ])),
  batch_size=batch_size_test, shuffle=True)
In [47]:
examples = enumerate(test_loader)
batch_idx, (example_data, example_targets) = next(examples)
In [48]:
example_data.shape
Out[48]:
torch.Size([1000, 1, 28, 28])
In [49]:
import matplotlib.pyplot as plt

fig = plt.figure()
for i in range(6):
  plt.subplot(2,3,i+1)
  plt.tight_layout()
  plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
  plt.title("Ground Truth: {}".format(example_targets[i]))
  plt.xticks([])
  plt.yticks([])
fig
Out[49]:
In [50]:
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
In [51]:
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x)
In [52]:
network = Net()
optimizer = optim.SGD(network.parameters(), lr=learning_rate,
                      momentum=momentum)
In [53]:
train_losses = []
train_counter = []
test_losses = []
test_counter = [i*len(train_loader.dataset) for i in range(n_epochs + 1)]
In [54]:
def train(epoch):
  network.train()
  for batch_idx, (data, target) in enumerate(train_loader):
    optimizer.zero_grad()
    output = network(data)
    loss = F.nll_loss(output, target)
    loss.backward()
    optimizer.step()
    if batch_idx % log_interval == 0:
      print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
        epoch, batch_idx * len(data), len(train_loader.dataset),
        100. * batch_idx / len(train_loader), loss.item()))
      train_losses.append(loss.item())
      train_counter.append(
        (batch_idx*64) + ((epoch-1)*len(train_loader.dataset)))
      torch.save(network.state_dict(), 'model.pth')
      torch.save(optimizer.state_dict(), 'optimizer.pth')
In [55]:
def test():
  network.eval()
  test_loss = 0
  correct = 0
  with torch.no_grad():
    for data, target in test_loader:
      output = network(data)
      test_loss += F.nll_loss(output, target, size_average=False).item()
      pred = output.data.max(1, keepdim=True)[1]
      correct += pred.eq(target.data.view_as(pred)).sum()
  test_loss /= len(test_loader.dataset)
  test_losses.append(test_loss)
  print('\nTest set: Avg. loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
    test_loss, correct, len(test_loader.dataset),
    100. * correct / len(test_loader.dataset)))
In [56]:
test()
for epoch in range(1, n_epochs + 1):
  train(epoch)
  test()

/tmp/ipykernel_69/161431047.py:17: UserWarning: Implicit dimension choice for log_softmax has been deprecated. Change the call to include dim=X as an argument.
  return F.log_softmax(x)

Test set: Avg. loss: 2.3096, Accuracy: 924/10000 (9%)

Train Epoch: 1 [0/60000 (0%)]	Loss: 2.313179
Train Epoch: 1 [640/60000 (1%)]	Loss: 2.300776
Train Epoch: 1 [1280/60000 (2%)]	Loss: 2.298872
Train Epoch: 1 [1920/60000 (3%)]	Loss: 2.276723
Train Epoch: 1 [2560/60000 (4%)]	Loss: 2.241102
Train Epoch: 1 [3200/60000 (5%)]	Loss: 2.248662
Train Epoch: 1 [3840/60000 (6%)]	Loss: 2.189547
Train Epoch: 1 [4480/60000 (7%)]	Loss: 2.146399
Train Epoch: 1 [5120/60000 (9%)]	Loss: 2.107701
Train Epoch: 1 [5760/60000 (10%)]	Loss: 2.011356
Train Epoch: 1 [6400/60000 (11%)]	Loss: 1.843857
Train Epoch: 1 [7040/60000 (12%)]	Loss: 1.758839
Train Epoch: 1 [7680/60000 (13%)]	Loss: 1.672644
Train Epoch: 1 [8320/60000 (14%)]	Loss: 1.687645
Train Epoch: 1 [8960/60000 (15%)]	Loss: 1.426628
Train Epoch: 1 [9600/60000 (16%)]	Loss: 1.284204
Train Epoch: 1 [10240/60000 (17%)]	Loss: 1.318824
Train Epoch: 1 [10880/60000 (18%)]	Loss: 1.445287
Train Epoch: 1 [11520/60000 (19%)]	Loss: 1.140686
Train Epoch: 1 [12160/60000 (20%)]	Loss: 1.152914
Train Epoch: 1 [12800/60000 (21%)]	Loss: 1.054201
Train Epoch: 1 [13440/60000 (22%)]	Loss: 0.900051
Train Epoch: 1 [14080/60000 (23%)]	Loss: 1.192117
Train Epoch: 1 [14720/60000 (25%)]	Loss: 0.944671
Train Epoch: 1 [15360/60000 (26%)]	Loss: 0.883142
Train Epoch: 1 [16000/60000 (27%)]	Loss: 0.846328
Train Epoch: 1 [16640/60000 (28%)]	Loss: 0.773332
Train Epoch: 1 [17280/60000 (29%)]	Loss: 0.760274
Train Epoch: 1 [17920/60000 (30%)]	Loss: 0.812324
Train Epoch: 1 [18560/60000 (31%)]	Loss: 0.892124
Train Epoch: 1 [19200/60000 (32%)]	Loss: 0.796882
Train Epoch: 1 [19840/60000 (33%)]	Loss: 0.910715
Train Epoch: 1 [20480/60000 (34%)]	Loss: 0.582793
Train Epoch: 1 [21120/60000 (35%)]	Loss: 0.691034
Train Epoch: 1 [21760/60000 (36%)]	Loss: 0.625940
Train Epoch: 1 [22400/60000 (37%)]	Loss: 0.770057
Train Epoch: 1 [23040/60000 (38%)]	Loss: 0.729386
Train Epoch: 1 [23680/60000 (39%)]	Loss: 0.633020
Train Epoch: 1 [24320/60000 (41%)]	Loss: 0.669564
Train Epoch: 1 [24960/60000 (42%)]	Loss: 0.837039
Train Epoch: 1 [25600/60000 (43%)]	Loss: 0.859584
Train Epoch: 1 [26240/60000 (44%)]	Loss: 0.607360
Train Epoch: 1 [26880/60000 (45%)]	Loss: 0.625770
Train Epoch: 1 [27520/60000 (46%)]	Loss: 0.684024
Train Epoch: 1 [28160/60000 (47%)]	Loss: 0.578418
Train Epoch: 1 [28800/60000 (48%)]	Loss: 0.557431
Train Epoch: 1 [29440/60000 (49%)]	Loss: 0.816638
Train Epoch: 1 [30080/60000 (50%)]	Loss: 0.750147
Train Epoch: 1 [30720/60000 (51%)]	Loss: 0.633136
Train Epoch: 1 [31360/60000 (52%)]	Loss: 0.644153
Train Epoch: 1 [32000/60000 (53%)]	Loss: 0.714072
Train Epoch: 1 [32640/60000 (54%)]	Loss: 0.448233
Train Epoch: 1 [33280/60000 (55%)]	Loss: 0.588381
Train Epoch: 1 [33920/60000 (57%)]	Loss: 0.571129
Train Epoch: 1 [34560/60000 (58%)]	Loss: 0.656939
Train Epoch: 1 [35200/60000 (59%)]	Loss: 0.569525
Train Epoch: 1 [35840/60000 (60%)]	Loss: 0.568231
Train Epoch: 1 [36480/60000 (61%)]	Loss: 0.633723
Train Epoch: 1 [37120/60000 (62%)]	Loss: 0.812953
Train Epoch: 1 [37760/60000 (63%)]	Loss: 0.577632
Train Epoch: 1 [38400/60000 (64%)]	Loss: 0.617762
Train Epoch: 1 [39040/60000 (65%)]	Loss: 0.542957
Train Epoch: 1 [39680/60000 (66%)]	Loss: 0.603089
Train Epoch: 1 [40320/60000 (67%)]	Loss: 0.319541
Train Epoch: 1 [40960/60000 (68%)]	Loss: 0.330860
Train Epoch: 1 [41600/60000 (69%)]	Loss: 0.382618
Train Epoch: 1 [42240/60000 (70%)]	Loss: 0.303939
Train Epoch: 1 [42880/60000 (71%)]	Loss: 0.414385
Train Epoch: 1 [43520/60000 (72%)]	Loss: 0.440848
Train Epoch: 1 [44160/60000 (74%)]	Loss: 0.455320
Train Epoch: 1 [44800/60000 (75%)]	Loss: 0.606288
Train Epoch: 1 [45440/60000 (76%)]	Loss: 0.542247
Train Epoch: 1 [46080/60000 (77%)]	Loss: 0.510507
Train Epoch: 1 [46720/60000 (78%)]	Loss: 0.376520
Train Epoch: 1 [47360/60000 (79%)]	Loss: 0.483785
Train Epoch: 1 [48000/60000 (80%)]	Loss: 0.553227
Train Epoch: 1 [48640/60000 (81%)]	Loss: 0.414906
Train Epoch: 1 [49280/60000 (82%)]	Loss: 0.389604
Train Epoch: 1 [49920/60000 (83%)]	Loss: 0.531326
Train Epoch: 1 [50560/60000 (84%)]	Loss: 0.428032
Train Epoch: 1 [51200/60000 (85%)]	Loss: 0.693468
Train Epoch: 1 [51840/60000 (86%)]	Loss: 0.507123
Train Epoch: 1 [52480/60000 (87%)]	Loss: 0.694401
Train Epoch: 1 [53120/60000 (88%)]	Loss: 0.534136
Train Epoch: 1 [53760/60000 (90%)]	Loss: 0.549806
Train Epoch: 1 [54400/60000 (91%)]	Loss: 0.645385
Train Epoch: 1 [55040/60000 (92%)]	Loss: 0.451940
Train Epoch: 1 [55680/60000 (93%)]	Loss: 0.560506
Train Epoch: 1 [56320/60000 (94%)]	Loss: 0.546273
Train Epoch: 1 [56960/60000 (95%)]	Loss: 0.355390
Train Epoch: 1 [57600/60000 (96%)]	Loss: 0.322211
Train Epoch: 1 [58240/60000 (97%)]	Loss: 0.294903
Train Epoch: 1 [58880/60000 (98%)]	Loss: 0.719042
Train Epoch: 1 [59520/60000 (99%)]	Loss: 0.378713

Test set: Avg. loss: 0.1848, Accuracy: 9445/10000 (94%)

Train Epoch: 2 [0/60000 (0%)]	Loss: 0.438718
Train Epoch: 2 [640/60000 (1%)]	Loss: 0.454982
Train Epoch: 2 [1280/60000 (2%)]	Loss: 0.500138
Train Epoch: 2 [1920/60000 (3%)]	Loss: 0.510807
Train Epoch: 2 [2560/60000 (4%)]	Loss: 0.300643
Train Epoch: 2 [3200/60000 (5%)]	Loss: 0.164536
Train Epoch: 2 [3840/60000 (6%)]	Loss: 0.486876
Train Epoch: 2 [4480/60000 (7%)]	Loss: 0.412689
Train Epoch: 2 [5120/60000 (9%)]	Loss: 0.380759
Train Epoch: 2 [5760/60000 (10%)]	Loss: 0.645167
Train Epoch: 2 [6400/60000 (11%)]	Loss: 0.579884
Train Epoch: 2 [7040/60000 (12%)]	Loss: 0.313967
Train Epoch: 2 [7680/60000 (13%)]	Loss: 0.424212
Train Epoch: 2 [8320/60000 (14%)]	Loss: 0.635905
Train Epoch: 2 [8960/60000 (15%)]	Loss: 0.486488
Train Epoch: 2 [9600/60000 (16%)]	Loss: 0.443411
Train Epoch: 2 [10240/60000 (17%)]	Loss: 0.312067
Train Epoch: 2 [10880/60000 (18%)]	Loss: 0.506013
Train Epoch: 2 [11520/60000 (19%)]	Loss: 0.324127
Train Epoch: 2 [12160/60000 (20%)]	Loss: 0.333954
Train Epoch: 2 [12800/60000 (21%)]	Loss: 0.471787
Train Epoch: 2 [13440/60000 (22%)]	Loss: 0.459535
Train Epoch: 2 [14080/60000 (23%)]	Loss: 0.475107
Train Epoch: 2 [14720/60000 (25%)]	Loss: 0.378559
Train Epoch: 2 [15360/60000 (26%)]	Loss: 0.205578
Train Epoch: 2 [16000/60000 (27%)]	Loss: 0.244983
Train Epoch: 2 [16640/60000 (28%)]	Loss: 0.366022
Train Epoch: 2 [17280/60000 (29%)]	Loss: 0.414595
Train Epoch: 2 [17920/60000 (30%)]	Loss: 0.328071
Train Epoch: 2 [18560/60000 (31%)]	Loss: 0.346349
Train Epoch: 2 [19200/60000 (32%)]	Loss: 0.502439
Train Epoch: 2 [19840/60000 (33%)]	Loss: 0.344948
Train Epoch: 2 [20480/60000 (34%)]	Loss: 0.372075
Train Epoch: 2 [21120/60000 (35%)]	Loss: 0.357003
Train Epoch: 2 [21760/60000 (36%)]	Loss: 0.491918
Train Epoch: 2 [22400/60000 (37%)]	Loss: 0.593303
Train Epoch: 2 [23040/60000 (38%)]	Loss: 0.503493
Train Epoch: 2 [23680/60000 (39%)]	Loss: 0.345845
Train Epoch: 2 [24320/60000 (41%)]	Loss: 0.395964
Train Epoch: 2 [24960/60000 (42%)]	Loss: 0.209819
Train Epoch: 2 [25600/60000 (43%)]	Loss: 0.201524
Train Epoch: 2 [26240/60000 (44%)]	Loss: 0.398361
Train Epoch: 2 [26880/60000 (45%)]	Loss: 0.294527
Train Epoch: 2 [27520/60000 (46%)]	Loss: 0.237990
Train Epoch: 2 [28160/60000 (47%)]	Loss: 0.176014
Train Epoch: 2 [28800/60000 (48%)]	Loss: 0.254591
Train Epoch: 2 [29440/60000 (49%)]	Loss: 0.515873
Train Epoch: 2 [30080/60000 (50%)]	Loss: 0.284150
Train Epoch: 2 [30720/60000 (51%)]	Loss: 0.316989
Train Epoch: 2 [31360/60000 (52%)]	Loss: 0.438148
Train Epoch: 2 [32000/60000 (53%)]	Loss: 0.677237
Train Epoch: 2 [32640/60000 (54%)]	Loss: 0.305531
Train Epoch: 2 [33280/60000 (55%)]	Loss: 0.289593
Train Epoch: 2 [33920/60000 (57%)]	Loss: 0.428519
Train Epoch: 2 [34560/60000 (58%)]	Loss: 0.394201
Train Epoch: 2 [35200/60000 (59%)]	Loss: 0.456256
Train Epoch: 2 [35840/60000 (60%)]	Loss: 0.453478
Train Epoch: 2 [36480/60000 (61%)]	Loss: 0.348346
Train Epoch: 2 [37120/60000 (62%)]	Loss: 0.488080
Train Epoch: 2 [37760/60000 (63%)]	Loss: 0.428703
Train Epoch: 2 [38400/60000 (64%)]	Loss: 0.273999
Train Epoch: 2 [39040/60000 (65%)]	Loss: 0.193952
Train Epoch: 2 [39680/60000 (66%)]	Loss: 0.415637
Train Epoch: 2 [40320/60000 (67%)]	Loss: 0.196386
Train Epoch: 2 [40960/60000 (68%)]	Loss: 0.488848
Train Epoch: 2 [41600/60000 (69%)]	Loss: 0.139882
Train Epoch: 2 [42240/60000 (70%)]	Loss: 0.353407
Train Epoch: 2 [42880/60000 (71%)]	Loss: 0.488660
Train Epoch: 2 [43520/60000 (72%)]	Loss: 0.275076
Train Epoch: 2 [44160/60000 (74%)]	Loss: 0.393356
Train Epoch: 2 [44800/60000 (75%)]	Loss: 0.236839
Train Epoch: 2 [45440/60000 (76%)]	Loss: 0.507139
Train Epoch: 2 [46080/60000 (77%)]	Loss: 0.295535
Train Epoch: 2 [46720/60000 (78%)]	Loss: 0.272118
Train Epoch: 2 [47360/60000 (79%)]	Loss: 0.197500
Train Epoch: 2 [48000/60000 (80%)]	Loss: 0.247274
Train Epoch: 2 [48640/60000 (81%)]	Loss: 0.369860
Train Epoch: 2 [49280/60000 (82%)]	Loss: 0.450887
Train Epoch: 2 [49920/60000 (83%)]	Loss: 0.275857
Train Epoch: 2 [50560/60000 (84%)]	Loss: 0.327943
Train Epoch: 2 [51200/60000 (85%)]	Loss: 0.444945
Train Epoch: 2 [51840/60000 (86%)]	Loss: 0.309726
Train Epoch: 2 [52480/60000 (87%)]	Loss: 0.347834
Train Epoch: 2 [53120/60000 (88%)]	Loss: 0.247479
Train Epoch: 2 [53760/60000 (90%)]	Loss: 0.236054
Train Epoch: 2 [54400/60000 (91%)]	Loss: 0.184396
Train Epoch: 2 [55040/60000 (92%)]	Loss: 0.480308
Train Epoch: 2 [55680/60000 (93%)]	Loss: 0.193984
Train Epoch: 2 [56320/60000 (94%)]	Loss: 0.366088
Train Epoch: 2 [56960/60000 (95%)]	Loss: 0.231687
Train Epoch: 2 [57600/60000 (96%)]	Loss: 0.571528
Train Epoch: 2 [58240/60000 (97%)]	Loss: 0.470542
Train Epoch: 2 [58880/60000 (98%)]	Loss: 0.282086
Train Epoch: 2 [59520/60000 (99%)]	Loss: 0.286647

Test set: Avg. loss: 0.1160, Accuracy: 9644/10000 (96%)

Train Epoch: 3 [0/60000 (0%)]	Loss: 0.205653
Train Epoch: 3 [640/60000 (1%)]	Loss: 0.245996
Train Epoch: 3 [1280/60000 (2%)]	Loss: 0.401412
Train Epoch: 3 [1920/60000 (3%)]	Loss: 0.261805
Train Epoch: 3 [2560/60000 (4%)]	Loss: 0.252195
Train Epoch: 3 [3200/60000 (5%)]	Loss: 0.323357
Train Epoch: 3 [3840/60000 (6%)]	Loss: 0.274522
Train Epoch: 3 [4480/60000 (7%)]	Loss: 0.494365
Train Epoch: 3 [5120/60000 (9%)]	Loss: 0.514789
Train Epoch: 3 [5760/60000 (10%)]	Loss: 0.396084
Train Epoch: 3 [6400/60000 (11%)]	Loss: 0.307430
Train Epoch: 3 [7040/60000 (12%)]	Loss: 0.338074
Train Epoch: 3 [7680/60000 (13%)]	Loss: 0.251372
Train Epoch: 3 [8320/60000 (14%)]	Loss: 0.160372
Train Epoch: 3 [8960/60000 (15%)]	Loss: 0.255050
Train Epoch: 3 [9600/60000 (16%)]	Loss: 0.225777
Train Epoch: 3 [10240/60000 (17%)]	Loss: 0.290729
Train Epoch: 3 [10880/60000 (18%)]	Loss: 0.294683
Train Epoch: 3 [11520/60000 (19%)]	Loss: 0.451191
Train Epoch: 3 [12160/60000 (20%)]	Loss: 0.223621
Train Epoch: 3 [12800/60000 (21%)]	Loss: 0.452726
Train Epoch: 3 [13440/60000 (22%)]	Loss: 0.345419
Train Epoch: 3 [14080/60000 (23%)]	Loss: 0.351051
Train Epoch: 3 [14720/60000 (25%)]	Loss: 0.323831
Train Epoch: 3 [15360/60000 (26%)]	Loss: 0.323383
Train Epoch: 3 [16000/60000 (27%)]	Loss: 0.802906
Train Epoch: 3 [16640/60000 (28%)]	Loss: 0.535793
Train Epoch: 3 [17280/60000 (29%)]	Loss: 0.444198
Train Epoch: 3 [17920/60000 (30%)]	Loss: 0.312039
Train Epoch: 3 [18560/60000 (31%)]	Loss: 0.306849
Train Epoch: 3 [19200/60000 (32%)]	Loss: 0.364884
Train Epoch: 3 [19840/60000 (33%)]	Loss: 0.370124
Train Epoch: 3 [20480/60000 (34%)]	Loss: 0.181933
Train Epoch: 3 [21120/60000 (35%)]	Loss: 0.179925
Train Epoch: 3 [21760/60000 (36%)]	Loss: 0.330862
Train Epoch: 3 [22400/60000 (37%)]	Loss: 0.237615
Train Epoch: 3 [23040/60000 (38%)]	Loss: 0.423070
Train Epoch: 3 [23680/60000 (39%)]	Loss: 0.204226
Train Epoch: 3 [24320/60000 (41%)]	Loss: 0.244979
Train Epoch: 3 [24960/60000 (42%)]	Loss: 0.173204
Train Epoch: 3 [25600/60000 (43%)]	Loss: 0.439161
Train Epoch: 3 [26240/60000 (44%)]	Loss: 0.190899
Train Epoch: 3 [26880/60000 (45%)]	Loss: 0.209164
Train Epoch: 3 [27520/60000 (46%)]	Loss: 0.212444
Train Epoch: 3 [28160/60000 (47%)]	Loss: 0.332085
Train Epoch: 3 [28800/60000 (48%)]	Loss: 0.481054
Train Epoch: 3 [29440/60000 (49%)]	Loss: 0.168669
Train Epoch: 3 [30080/60000 (50%)]	Loss: 0.265230
Train Epoch: 3 [30720/60000 (51%)]	Loss: 0.234366
Train Epoch: 3 [31360/60000 (52%)]	Loss: 0.214331
Train Epoch: 3 [32000/60000 (53%)]	Loss: 0.454029
Train Epoch: 3 [32640/60000 (54%)]	Loss: 0.340459
Train Epoch: 3 [33280/60000 (55%)]	Loss: 0.172732
Train Epoch: 3 [33920/60000 (57%)]	Loss: 0.459111
Train Epoch: 3 [34560/60000 (58%)]	Loss: 0.258277
Train Epoch: 3 [35200/60000 (59%)]	Loss: 0.344975
Train Epoch: 3 [35840/60000 (60%)]	Loss: 0.464806
Train Epoch: 3 [36480/60000 (61%)]	Loss: 0.184675
Train Epoch: 3 [37120/60000 (62%)]	Loss: 0.213830
Train Epoch: 3 [37760/60000 (63%)]	Loss: 0.302575
Train Epoch: 3 [38400/60000 (64%)]	Loss: 0.461729
Train Epoch: 3 [39040/60000 (65%)]	Loss: 0.337852
Train Epoch: 3 [39680/60000 (66%)]	Loss: 0.199724
Train Epoch: 3 [40320/60000 (67%)]	Loss: 0.231916
Train Epoch: 3 [40960/60000 (68%)]	Loss: 0.451854
Train Epoch: 3 [41600/60000 (69%)]	Loss: 0.330580
Train Epoch: 3 [42240/60000 (70%)]	Loss: 0.387095
Train Epoch: 3 [42880/60000 (71%)]	Loss: 0.180348
Train Epoch: 3 [43520/60000 (72%)]	Loss: 0.248029
Train Epoch: 3 [44160/60000 (74%)]	Loss: 0.295474
Train Epoch: 3 [44800/60000 (75%)]	Loss: 0.180743
Train Epoch: 3 [45440/60000 (76%)]	Loss: 0.243530
Train Epoch: 3 [46080/60000 (77%)]	Loss: 0.343313
Train Epoch: 3 [46720/60000 (78%)]	Loss: 0.200602
Train Epoch: 3 [47360/60000 (79%)]	Loss: 0.398185
Train Epoch: 3 [48000/60000 (80%)]	Loss: 0.333912
Train Epoch: 3 [48640/60000 (81%)]	Loss: 0.204814
Train Epoch: 3 [49280/60000 (82%)]	Loss: 0.229588
Train Epoch: 3 [49920/60000 (83%)]	Loss: 0.499256
Train Epoch: 3 [50560/60000 (84%)]	Loss: 0.401688
Train Epoch: 3 [51200/60000 (85%)]	Loss: 0.422475
Train Epoch: 3 [51840/60000 (86%)]	Loss: 0.233126
Train Epoch: 3 [52480/60000 (87%)]	Loss: 0.495838
Train Epoch: 3 [53120/60000 (88%)]	Loss: 0.232005
Train Epoch: 3 [53760/60000 (90%)]	Loss: 0.191488
Train Epoch: 3 [54400/60000 (91%)]	Loss: 0.400002
Train Epoch: 3 [55040/60000 (92%)]	Loss: 0.310593
Train Epoch: 3 [55680/60000 (93%)]	Loss: 0.177988
Train Epoch: 3 [56320/60000 (94%)]	Loss: 0.282546
Train Epoch: 3 [56960/60000 (95%)]	Loss: 0.325411
Train Epoch: 3 [57600/60000 (96%)]	Loss: 0.552869
Train Epoch: 3 [58240/60000 (97%)]	Loss: 0.370660
Train Epoch: 3 [58880/60000 (98%)]	Loss: 0.117798
Train Epoch: 3 [59520/60000 (99%)]	Loss: 0.301632

Test set: Avg. loss: 0.0995, Accuracy: 9679/10000 (97%)
In [65]:
fig = plt.figure()
plt.plot(train_counter, train_losses, color='blue')
plt.scatter(test_counter, test_losses, color='red')
plt.legend(['Train Loss', 'Test Loss'], loc='upper right')
plt.xlabel('number of training examples seen')
plt.ylabel('negative log likelihood loss')
fig
Out[65]:
In [62]:
len(test_counter), len(test_losses)
Out[62]:
(4, 4)
In [66]:
with torch.no_grad():
  output = network(example_data)

/tmp/ipykernel_69/161431047.py:17: UserWarning: Implicit dimension choice for log_softmax has been deprecated. Change the call to include dim=X as an argument.
  return F.log_softmax(x)
In [67]:
fig = plt.figure()
for i in range(6):
  plt.subplot(2,3,i+1)
  plt.tight_layout()
  plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
  plt.title("Prediction: {}".format(
    output.data.max(1, keepdim=True)[1][i].item()))
  plt.xticks([])
  plt.yticks([])
fig
Out[67]:
In [ ]: