Train loss is low

Hi all,

I am trying to train vgg13 (pre train model) to classify the images. I put my classified
classifier = nn.Sequential(OrderedDict([
(‘fc1’, nn.Linear(25088, 5000)),
(‘relu’, nn.ReLU()),
(‘dropout’, nn.Dropout(0.2)),
(‘fc2’, nn.Linear(5000, 102)),
(‘output’, nn.LogSoftmax(dim=1))

when I start training this model

Epoch: 1/10… Training Loss: 0.0608 Test Loss:1.786 Test Accurracy:0.564
Epoch: 1/10… Training Loss: 0.0408 Test Loss:1.032 Test Accurracy:0.730

As you can see the Training Loss is too low and I am not able to reason why. Any clue?