Я пишу следующий код, но я получил эту ошибку:
TypeError: backward() got an unexpected keyword argument 'retain_variables'
Мой код:
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()