I’m trying to train a pre-trained Inception v3 model for my task, which gives as input 178x178 images. It has 5 possible classes so I changed the fully-connected layer to have 5 output feature. My code is the following:
# Pre-trained models
model = models.inception_v3(pretrained=True)
### ResNet or Inception
classifier_input = model.fc.in_features
num_labels = 5
# Replace default classifier with new classifier
model.fc = nn.Linear(classifier_input, num_labels)
model.cuda()
However, I’m getting the following error: RuntimeError: Calculated padded input size per channel: (2 x 2). Kernel size: (5 x 5). Kernel size can’t be greater than actual input size.
I’m not sure what’s the problem as I adopted a similar strategy when training a pre-trained VGG and ResNet.
The output of inception will now return an InceptionOutput, which will contain the .logits and .aux_logits, if specified.
If you don’t need the aus_logits, just use output.logits in your further processing.
I tried doing this and have researched more but have not been successful.
My training code is the following:
def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
since = time.time()
best_model_wts = copy.deepcopy(model.state_dict())
best_acc = 0.0
best_epoch = -1
for epoch in range(num_epochs):
print('Epoch {}/{}'.format(epoch, num_epochs - 1))
print('-' * 10)
# Each epoch has a training and validation phase
for phase in ['train', 'val']:
if phase == 'train':
model.train() # Set model to training mode
else:
model.eval() # Set model to evaluate mode
running_loss = 0.0
running_corrects = 0
# Iterate over data.
for inputs, labels in dataloaders[phase]:
if train_on_gpu:
inputs, labels = inputs.cuda(), labels.cuda()
# zero the parameter gradients
optimizer.zero_grad()
# forward
# track history if only in train
with torch.set_grad_enabled(phase == 'train'):
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
loss = criterion(outputs, labels)
# backward + optimize only if in training phase
if phase == 'train':
loss.backward()
optimizer.step()
# statistics
running_loss += loss.item() * inputs.size(0)
running_corrects += torch.sum(preds == labels.data)
if phase == 'train':
scheduler.step()
# deep copy the model
if phase == 'val' and epoch_acc > best_acc:
best_acc = epoch_acc
best_model_wts = copy.deepcopy(model.state_dict())
best_epoch = epoch
print()
time_elapsed = time.time() - since
print('Training complete in {:.0f}m {:.0f}s'.format(
time_elapsed // 60, time_elapsed % 60))
print('Best val Acc: {:4f}'.format(best_acc))
print('Best epoch: {:4f}'.format(best_epoch))
# load best model weights
model.load_state_dict(best_model_wts)
return model