是否可以在不重新训练模型的情况下解决 TypeError: argument 'input' (position 1) must be Tensor error？

如何解决是否可以在不重新训练模型的情况下解决 TypeError: argument 'input' (position 1) must be Tensor error？

我在 PyTorch 中制作了一个模型，用于 openAI Gym 环境。我是通过以下方式做到的：

class Policy(nn.Module):
    def __init__(self,s_size=8,h_size=16,a_size=4):
        super(Policy,self).__init__()
        self.fc1 = nn.Linear(s_size,h_size)
        self.fc2 = nn.Linear(h_size,32)
        self.fc3 = nn.Linear(32,64)
        self.fc4 = nn.Linear(64,a_size)

    def forward(self,x):
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = F.relu(self.fc3(x))
        x = self.fc4(x)
        return F.softmax(x,dim=1 )
    
    def act(self,state):
        state = torch.from_numpy(state).float().unsqueeze(0).to(device)
        probs = self.forward(state).cpu()
        m = Categorical(probs)
        action = m.sample()
        return action.item(),m.log_prob(action)

然后我将它的状态保存在字典中并按如下方式使用它：

env = gym.make('LunarLander-v2')

policy = Policy().to(torch.device('cpu'))
policy.load_state_dict(torch.load('best_params_cloud.ckpt',map_location='cpu'))
policy.eval()
ims = []
rewards = []
state = env.reset()
for step in range(STEPS):
    img = env.render(mode='rgb_array')
    action,log_prob = policy(state)
        # print(action)
    state,reward,done,i_ = env.step(action)
    rewards.append(reward)
    # print(reward,done)
    cv2_im_rgb = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
    pil_im = Image.fromarray(cv2_im_rgb)

    draw = ImageDraw.Draw(pil_im)

    # Choose a font
    font = ImageFont.truetype("Roboto-Regular.ttf",20)

    # Draw the text
    draw.text((0,0),f"Step: {step} Action : {action} Reward: {int(reward)} Total Rewards: {int(np.sum(rewards))} done: {done}",font=font,fill="#FDFEFE")

    # Save the image
    img = cv2.cvtColor(np.array(pil_im),cv2.COLOR_RGB2BGR)
    im = plt.imshow(img,animated=True)
    ims.append([im])
    if done:
        env.close()


                
        
        break

Writer = animation.writers['pillow']
writer = Writer(fps=15,Metadata=dict(artist='Me'),bitrate=1800)
im_ani = animation.ArtistAnimation(fig,ims,interval=50,repeat_delay=3000,blit=True)
im_ani.save('ll_train1.gif',writer=writer)

但这会返回错误：

TypeError                                 Traceback (most recent call last)
<ipython-input-3-da32222edde2> in <module>
      9 for step in range(STEPS):
     10     img = env.render(mode='rgb_array')
---> 11     action,log_prob = policy(state)
     12         # print(action)
     13     state,i_ = env.step(action)

~\anaconda3\lib\site-packages\torch\nn\modules\module.py in _call_impl(self,*input,**kwargs)
    887             result = self._slow_forward(*input,**kwargs)
    888         else:
--> 889             result = self.forward(*input,**kwargs)
    890         for hook in itertools.chain(
    891                 _global_forward_hooks.values(),<ipython-input-2-66d42ebb791e> in forward(self,x)
     33 
     34     def forward(self,x):
---> 35         x = F.relu(self.fc1(x))
     36         x = F.relu(self.fc2(x))
     37         x = F.relu(self.fc3(x))

~\anaconda3\lib\site-packages\torch\nn\modules\module.py in _call_impl(self,~\anaconda3\lib\site-packages\torch\nn\modules\linear.py in forward(self,input)
     92 
     93     def forward(self,input: Tensor) -> Tensor:
---> 94         return F.linear(input,self.weight,self.bias)
     95 
     96     def extra_repr(self) -> str:

~\anaconda3\lib\site-packages\torch\nn\functional.py in linear(input,weight,bias)
   1751     if has_torch_function_variadic(input,weight):
   1752         return handle_torch_function(linear,(input,weight),input,bias=bias)
-> 1753     return torch._C._nn.linear(input,bias)
   1754 
   1755 

TypeError: linear(): argument 'input' (position 1) must be Tensor,not numpy.ndarray

我尝试通过添加以下代码行来更改转发功能：

def forward(self,x):
        x = torch.tensor(x,dtype=torch.float32,device=DEVICE).unsqueeze(0) //Added this line
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = F.relu(self.fc3(x))
        x = self.fc4(x)
        return F.softmax(x,dim=1 )

但这也会返回错误：ValueError: not enough values to unpack (expected 2,got 1)

该策略花费了大量时间来训练，而我试图避免对其进行再训练，是否有解决方法使其无需再训练即可运行？

解决方法

此错误与您的模型无关。 forward 函数只返回概率分布，但您需要的是动作和对应的概率（Policy.act 的输出）。

更改您的代码

for step in range(STEPS):
    img = env.render(mode='rgb_array')

    # This line causes the error.
    action,log_prob = policy(state)

到

for step in range(STEPS):
    img = env.render(mode='rgb_array')

    # This line causes the error.
    action,log_prob = policy.act(state)