我在尝试用PyTorch创建一个AI时遇到了这个错误：

RuntimeError: gather_out_cpu(): Expected dtype int64 for index

这是我的函数：

def learn(self, batch_state, batch_next_state, batch_reward, batch_action):    outputs = self.model(batch_state).gather(1, batch_action.unsqueeze(1)).squeeze(1)    next_outputs = self.model(batch_next_state).detach().max(1)[0]    target = self.gamma * next_outputs + batch_reward    td_loss = F.smooth_l1_loss(outputs, target)    self.optimizer.zero_grad()    td_loss.backward(retain_variables = True)    self.optimizer.step()

回答：

您需要在将batch_action张量传递给torch.gather之前更改其数据类型。

def learn(...):    batch_action = batch_action.type(torch.int64)     outputs = ...    ...# oroutputs = self.model(batch_state).gather(1, batch_action.type(torch.int64).unsqueeze(1)).squeeze(1)