How I should change the keywords in the training loop to get accuracy ?
for epoch in range(EPOCHS):
print('Epoch ', epoch,'/',EPOCHS-1)
print('-'*15)
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 = 0.0
# Iterate over data.
for inputs,labels in dataloaders[phase]:
inputs = inputs.to(device, dtype=torch.float)
labels = labels.to(device, dtype=torch.float)
# zero the parameter gradients
optimizer.zero_grad()
with torch.set_grad_enabled(phase == 'train'):
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
loss = loss_fn(outputs,labels)
# we backpropagate to set our learning parameters only in training mode
if phase == 'train':
loss.backward()
optimizer.step()
running_loss += loss.item() * inputs.size(0)
if phase == 'val':
running_corrects += torch.sum(preds == torch.argmax(labels.data, dim=1))
# scheduler for weight decay
if phase == 'train':
scheduler.step()
epoch_loss = running_loss / float(dataset_sizes[phase])
xm.master_print('{} Loss: {:.4f}'.format(phase, epoch_loss))
else:
epoch_loss = running_loss / float(dataset_sizes[phase])
epoch_acc = running_corrects / float(dataset_sizes[phase])
xm.master_print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
I am runnig this notebook in integration with torch xla with Tpu …if I remove that line it is giving me error .
Can you try copy and edit this notebook
I changed the code a little bit and it started working . But the loss is still to high ; What should I make changes in model class to improve the accuracy ? I am Training 51000 images with a batch size of 256 sized at 224 x 224 , for 25 epochs. Notebook Link
import torch.nn as nn
import torch.nn.functional as F
class EfficientNet_b0(nn.Module):
def __init__(self):
super(EfficientNet_b0, self).__init__()
self.model = efficientnet_pytorch.EfficientNet.from_pretrained('efficientnet-b0')
self.classifier_layer = nn.Sequential(
nn.Linear(1280 , 512),
nn.BatchNorm1d(512),
nn.Dropout(0.2),
nn.Linear(512 , 256),
nn.Linear(256 , 104)
)
def forward(self, inputs):
x = self.model.extract_features(inputs)
# Pooling and final linear layer
x = self.model._avg_pooling(x)
x = x.flatten(start_dim=1)
x = self.model._dropout(x)
x = self.classifier_layer(x)
return x
model = EfficientNet_b0()
model = model.to(device)
You might want to add activation functions into classifier_layer. If that doesn’t help, try to overfit a small data sample (e.g. just 10 samples) by playing around with some hyperparameters. Once this is done you could try to scale up the use case again.
I changed the model after your comments but it’s not heling me out rather the accuracy is decreasing
Need help in how to increase the accuracy in retrained mdels using pytorch ? Any good references will also help! Notebook Link
class EfficientNet_b0(nn.Module):
def __init__(self):
super(EfficientNet_b0, self).__init__()
self.model = efficientnet_pytorch.EfficientNet.from_pretrained('efficientnet-b0')
self.dense_layer_1 = nn.Linear(1280 , 512)
self.batchNorm_layer = nn.BatchNorm1d(512)
self.droput_layer = nn.Dropout(0.2)
self.dense_layer_2 = nn.Linear(512,216)
self.dense_layer_3 = nn.Linear(216,104)
def forward(self, inputs):
x = self.model.extract_features(inputs)
# Pooling and final linear layer
x = self.model._avg_pooling(x)
x = x.flatten(start_dim=1)
x = self.model._dropout(x)
x = self.dense_layer_1(x)
x = self.batchNorm_layer(x)
x = self.droput_layer(x)
x = self.dense_layer_2(x)
x = torch.relu(x)
x = self.dense_layer_3(x)
x = torch.log_softmax(x , dim =1)
return x
model = EfficientNet_b0()
