RuntimeError：设备类型为 cuda 的预期对象，但在调用 _th_index_select 站点：stackoverflow.com 时获得了参数 #3 'index' 的设备类型 cpu答案

【问题标题】：RuntimeError: Expected object of device type cuda but got device type cpu for argument #3 'index' in call to _th_index_select site:stackoverflow.comRuntimeError：设备类型为 cuda 的预期对象，但在调用 _th_index_select 站点：stackoverflow.com 时获得了参数 #3 'index' 的设备类型 cpu
【发布时间】：2020-06-01 13:36:16
【问题描述】：

我在 SMILE 数据集上使用了 bert。我写了以下代码，你能指导我哪里出错了。我已经编写了正确评估的培训代码，但是当我尝试运行评估代码进行验证时，它给出了错误。我试图将参数直接传递给 cuda。我仍然面临这个问题

'''

def evaluate(dataloader_val):

  print("in evaluate")
  model.eval()

  loss_val_total = 0
  predictions, true_value = [],[]

  for batch in dataloader_val:

    print("in for loop of dataloader")
    barch = tuple(b.to(device) for b in batch)

    inputs = {
               'input_ids':  batch[0],
                'attention_mask': batch[1],
                 'labels' : batch[2],
    }

    with torch.no_grad():
      outputs = model(**inputs)

    loss = outputs[0]
    logits = outputs[1]
    loss_val_total += loss.item()

    print("before logit")

    logits = logits.to(device)
    print("in the for batch evaluate: ",logits)
    label_ids = inputs['labels'].to(device)
    true_vals.append(label_ids)

  loss_val_avg = loss_val_total/len(dataloader_val)

  predictions = np.concatenate(predictions, axis = 0)
  true_vals = np.concatenate(true_vals,axis = 0)

  return loss_val_avg, predictions, true_vals
'''

另一个函数是

'''
for epoch in tqdm(range(1, epochs+1)):
  model.train()

  loss_train_total = 0

  progress_bar = tqdm(dataloader_train,
                      desc = 'Epoch {:1d}'.format(epoch),
                      leave = False,
                      disable = False)
  for batch in progress_bar:

    model.zero_grad()

    batch = tuple(b.to(device) for b in batch)

    inputs = {

            'input_ids'      : batch[0],
            'attention_mask' : batch[1],
            'labels'         : batch[2]
    }

    outputs = model(**inputs)

    loss = outputs[0]
    loss_train_total += loss.item()
    loss.backward()

    torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)

    optimizer.step()
    scheduler.step()

    progress_bar.set_postfix({'training_loss' : '{:.3f}'.format(loss.item()/len(batch))})

  torch.save(model.state_dict(), f'/content/drive/My Drive/Bert/Coursera/SMILE/Bert_ft_epoch{epoch}.model')

  tqdm.write(f'\n Epoch {epoch}')

  loss_train_avg = loss_train_total / len(dataloader_train)

  tqdm.write(f'Training Loss: {loss_train_avg}')

  val_loss, predictions, true_vals = evaluate(dataloader_val)
  val_f1 = f1_score_func(predictions, true_vals)
  tqdm.write(f'Validation loss : {val_loss}')
  tqdm.write(f'F1 score(weighted): {val_f1}')
'''

【问题讨论】：

标签： python pytorch huggingface-transformers

【解决方案1】：

您的评估函数中有错字：

barch = tuple(b.to(device) for b in batch)

您将 gpu 数据分配给 barch 而不是 batch。

【讨论】：