UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and

若在Pytorch训练网络中出现一下错误警告的解决办法如下：

UserWarning: Detected call of `lr_scheduler.step()` before `optimizer.step()`. In PyTorch 1.1.0 and later, you should call them in the opposite order: `optimizer.step()` before `lr_scheduler.step()`.  Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
  "https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)

错误原因在“optimizer.step（）”之前检测到对“lr_scheduler.step（）”的调用。在 PyTorch 1.1.0 及更高版本中，您应该以相反的顺序调用它们：'optimizer.step（）' 在 'lr_scheduler.step（）' 之前。 如果不这样做，将导致 PyTorch 跳过学习率计划的第一个值。

例如代码修改前：

for epoch in range( 1 + opt.continueEpochs, opt.nEpochs +1 + opt.continueEpochs):
    
    print("Training...")
    
    scheduler.step()

    epoch_loss = 0

    psnr_list = []
    for iteration, inputs in enumerate(train_dataloader,1):

        haze, gt = Variable(inputs['hazy_image']), Variable(inputs['clear_image'])
        
        haze = haze.cuda()
        #print(haze.shape)
        gt = gt.cuda()

        # --- Zero the parameter gradients --- #
        optimizer.zero_grad()

        # --- Forward + Backward + Optimize --- #
        model.train()
        dehaze = model(haze)
        #print(dehaze.size())
        #MSE_loss = MSELoss(dehaze, gt)
        #msssim_loss_ =1 -msssim_loss(dehaze, gt, normalize=True)
        Loss1 = loss_function_at(dehaze,gt)
        perceptual_loss = loss_network(dehaze,gt)
        #EDGE_loss = edge_loss(dehaze, gt, device)
        #ContrastLoss = ContrastLoss(dehaze)
        Loss = Loss1 +0.01*perceptual_loss# + 0.2*msssim_loss_
        epoch_loss +=Loss
        Loss.backward()
        optimizer.step()

代码修改后消除警告

for epoch in range( 1 + opt.continueEpochs, opt.nEpochs +1 + opt.continueEpochs):
    
    print("Training...")
    

    epoch_loss = 0
    psnr_list = []
    for iteration, inputs in enumerate(train_dataloader,1):

        haze, gt = Variable(inputs['hazy_image']), Variable(inputs['clear_image'])
        
        haze = haze.to(device) 
        gt = gt.to(device)

        # --- Zero the parameter gradients --- #
        optimizer.zero_grad()

        # --- Forward + Backward + Optimize --- #
        model.train()
        dehaze = model(haze)
        #print(dehaze.size())
        #MSE_loss = MSELoss(dehaze, gt)
        #msssim_loss_ =1 -msssim_loss(dehaze, gt, normalize=True)
        Loss1 = loss_function_at(dehaze,gt)
        perceptual_loss = loss_network(dehaze,gt)
        #EDGE_loss = edge_loss(dehaze, gt, device)
        #ContrastLoss = ContrastLoss(dehaze)
        Loss = Loss1 +0.01*perceptual_loss# + 0.2*msssim_loss_
        epoch_loss +=Loss
        Loss.backward()
        optimizer.step()
        scheduler.step()   

 消除警告