输出评估损失后，每n批次，而不是时代与pytorch

Question

我想在每个n批次之后输出它，而不是在每个纪元打印评估损失。

我每个时代大约有15万批。我想每50000批输出一次评估损失。

这可能吗？我正在使用pytorch和huggingface的预训练bert模型。

我的列车循环：

best_valid_loss = float('inf')
train_losses=[]
valid_losses=[]

for epoch in range(params['epochs']):
     
      print('\n Epoch {:} / {:}'.format(epoch + 1, params['epochs']))
      
      #train model
      train_loss = train(scheduler, optimizer)
      
      #evaluate model
      valid_loss = evaluate()
      
      #save the best model
      if valid_loss < best_valid_loss:
          best_valid_loss = valid_loss
          torch.save(model.state_dict(), model_file)
      
      # append training and validation loss
      train_losses.append(train_loss)
      valid_losses.append(valid_loss)
      
      print(f'\nTraining Loss: {train_loss:.3f}')
      print(f'Validation Loss: {valid_loss:.3f}')

Answer 1

如果要跳过例如某些列表的n元素，可以使用enumerate：

n = 50000

for i,epoch in enumerate(some_list):
     if i%n == 0:
     
           print('\n Epoch {:} / {:}'.format(epoch + 1, params['epochs']))
           ...

  

但在您的情况下，您只能使用一个附加条件：

n = 50000

for epoch in range(params['epochs']):
     #train model
     train_loss = train(scheduler, optimizer)
     
     #evaluate model
     valid_loss = evaluate()
  
     #save the best model
     if valid_loss < best_valid_loss:
           best_valid_loss = valid_loss
           torch.save(model.state_dict(), model_file)
  
     # append training and validation loss
     train_losses.append(train_loss)
     valid_losses.append(valid_loss)
  
     ##
     # print every n cycles 
     if epoch%n == 0:
           print('\n Epoch {:} / {:}'.format(epoch + 1, params['epochs']))

           print(f'\nTraining Loss: {train_loss:.3f}')
           print(f'Validation Loss: {valid_loss:.3f}')