RuntimeError: Error(s) in loading state_dict for Actor - torch.load()
RuntimeError: Error(s) in loading state_dict for Actor - torch.load()
我在 open ai gym 中创建了一个自定义环境，我在加载重量时遇到错误有人可以帮助我解决这个问题吗？我正在自定义环境中训练 TD3 网络，我已经成功训练，但在推理时我遇到了这个问题
class Actor(nn.Module):
def __init__(self, state_dim, action_dim, max_action):
super(Actor, self).__init__()
self.layer_1 = nn.Linear(state_dim, 400)
self.layer_2 = nn.Linear(400, 300)
self.layer_3 = nn.Linear(300, action_dim)
self.max_action = max_action
def forward(self, x):
x = F.relu(self.layer_1(x))
x = F.relu(self.layer_2(x))
x = self.max_action * torch.tanh(self.layer_3(x))
return x
class Critic(nn.Module):
def __init__(self, state_dim, action_dim):
super(Critic, self).__init__()
# Defining the first Critic neural network
self.layer_1 = nn.Linear(state_dim + action_dim, 400)
self.layer_2 = nn.Linear(400, 300)
self.layer_3 = nn.Linear(300, 1)
# Defining the second Critic neural network
self.layer_4 = nn.Linear(state_dim + action_dim, 400)
self.layer_5 = nn.Linear(400, 300)
self.layer_6 = nn.Linear(300, 1)
def forward(self, x, u):
xu = torch.cat([x, u], 1)
# Forward-Propagation on the first Critic Neural Network
x1 = F.relu(self.layer_1(xu))
x1 = F.relu(self.layer_2(x1))
x1 = self.layer_3(x1)
# Forward-Propagation on the second Critic Neural Network
x2 = F.relu(self.layer_4(xu))
x2 = F.relu(self.layer_5(x2))
x2 = self.layer_6(x2)
return x1, x2
def Q1(self, x, u):
xu = torch.cat([x, u], 1)
x1 = F.relu(self.layer_1(xu))
x1 = F.relu(self.layer_2(x1))
x1 = self.layer_3(x1)
return x1
# Selecting the device (CPU or GPU)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Building the whole Training Process into a class
class TD3(object):
def __init__(self, state_dim, action_dim, max_action):
self.actor = Actor(state_dim, action_dim, max_action).to(device)
self.actor_target = Actor(state_dim, action_dim, max_action).to(device)
self.actor_target.load_state_dict(self.actor.state_dict())
self.actor_optimizer = torch.optim.Adam(self.actor.parameters())
self.critic = Critic(state_dim, action_dim).to(device)
self.critic_target = Critic(state_dim, action_dim).to(device)
self.critic_target.load_state_dict(self.critic.state_dict())
self.critic_optimizer = torch.optim.Adam(self.critic.parameters())
self.max_action = max_action
def select_action(self, state):
state = torch.Tensor(state.reshape(1, -1)).to(device)
return self.actor(state).cpu().data.numpy().flatten()
def train(self, replay_buffer, iterations, batch_size=100, discount=0.99, tau=0.005, policy_noise=0.2, noise_clip=0.5, policy_freq=2):
for it in range(iterations):
# Step 4: We sample a batch of transitions (s, s’, a, r) from the memory
batch_states, batch_next_states, batch_actions, batch_rewards, batch_dones = replay_buffer.sample(batch_size)
state = torch.Tensor(batch_states).to(device)
next_state = torch.Tensor(batch_next_states).to(device)
action = torch.Tensor(batch_actions).to(device)
reward = torch.Tensor(batch_rewards).to(device)
done = torch.Tensor(batch_dones).to(device)
# Step 5: From the next state s’, the Actor target plays the next action a’
next_action = self.actor_target(next_state)
# Step 6: We add Gaussian noise to this next action a’ and we clamp it in a range of values supported by the environment
noise = torch.Tensor(batch_actions).data.normal_(0, policy_noise).to(device)
noise = noise.clamp(-noise_clip, noise_clip)
next_action = (next_action + noise).clamp(-self.max_action, self.max_action)
# Step 7: The two Critic targets take each the couple (s’, a’) as input and return two Q-values Qt1(s’,a’) and Qt2(s’,a’) as outputs
target_Q1, target_Q2 = self.critic_target(next_state, next_action)
# Step 8: We keep the minimum of these two Q-values: min(Qt1, Qt2)
target_Q = torch.min(target_Q1, target_Q2)
# Step 9: We get the final target of the two Critic models, which is: Qt = r + γ * min(Qt1, Qt2), where γ is the discount factor
target_Q = reward + ((1 - done) * discount * target_Q).detach()
# Step 10: The two Critic models take each the couple (s, a) as input and return two Q-values Q1(s,a) and Q2(s,a) as outputs
current_Q1, current_Q2 = self.critic(state, action)
# Step 11: We compute the loss coming from the two Critic models: Critic Loss = MSE_Loss(Q1(s,a), Qt) + MSE_Loss(Q2(s,a), Qt)
critic_loss = F.mse_loss(current_Q1, target_Q) + F.mse_loss(current_Q2, target_Q)
# Step 12: We backpropagate this Critic loss and update the parameters of the two Critic models with a SGD optimizer
self.critic_optimizer.zero_grad()
critic_loss.backward()
self.critic_optimizer.step()
# Step 13: Once every two iterations, we update our Actor model by performing gradient ascent on the output of the first Critic model
if it % policy_freq == 0:
actor_loss = -self.critic.Q1(state, self.actor(state)).mean()
self.actor_optimizer.zero_grad()
actor_loss.backward()
self.actor_optimizer.step()
# Step 14: Still once every two iterations, we update the weights of the Actor target by polyak averaging
for param, target_param in zip(self.critic.parameters(), self.critic_target.parameters()):
target_param.data.copy_(tau * param.data + (1 - tau) * target_param.data)
# Step 15: Still once every two iterations, we update the weights of the Critic target by polyak averaging
for param, target_param in zip(self.actor.parameters(), self.actor_target.parameters()):
target_param.data.copy_(tau * param.data + (1 - tau) * target_param.data)
# Making a save method to save a trained model
def save(self, filename, directory):
torch.save(self.actor.state_dict(), '%s/%s_actor.pth' % (directory, filename))
torch.save(self.critic.state_dict(), '%s/%s_critic.pth' % (directory, filename))
# Making a load method to load a pre-trained model
def load(self, filename, directory):
self.actor.load_state_dict(torch.load('%s/%s_actor.pth' % (directory, filename)))
self.critic.load_state_dict(torch.load('%s/%s_critic.pth' % (directory, filename)))
def evaluate_policy(policy, eval_episodes=10):
avg_reward = 0.
for _ in range(eval_episodes):
obs = env.reset()
done = False
while not done:
action = policy.select_action(np.array(obs))
obs, reward, done, _ = env.step(action)
avg_reward += reward
avg_reward /= eval_episodes
print ("---------------------------------------")
print ("Average Reward over the Evaluation Step: %f" % (avg_reward))
print ("---------------------------------------")
return avg_reward
env_name = "Pygame-v0"
seed = 0
file_name = "%s_%s_%s" % ("TD3", env_name, str(seed))
print ("---------------------------------------")
print ("Settings: %s" % (file_name))
print ("---------------------------------------")
eval_episodes = 10
save_env_vid = True
env = gym.make(env_name)
max_episode_steps = env._max_episode_steps
if save_env_vid:
env = wrappers.Monitor(env, monitor_dir, force = True)
env.reset()
env.seed(seed)
torch.manual_seed(seed)
np.random.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
max_action = float(env.action_space.high[0])
policy = TD3(state_dim, action_dim, max_action)
#policy.load(file_name, './pytorch_models/')
policy.load(file_name,"/content/gdrive/My Drive/reinforce/gym_game/pytorch_models")
_ = evaluate_policy(policy, eval_episodes=eval_episodes)
回溯：
我在为演员模型加载 state_dict 时遇到运行时错误。我搜索了 google 但找不到类似的问题。
RuntimeError: Error(s) in loading state_dict for Actor:
Missing key(s) in state_dict: "layer_1.weight", "layer_1.bias", "layer_2.weight", "layer_2.bias", "layer_3.weight", "layer_3.bias".
Unexpected key(s) in state_dict: "encoder.0.weight", "encoder.0.bias", "encoder.2.weight", "encoder.2.bias", "encoder.2.running_mean", "encoder.2.running_var", "encoder.2.num_batches_tracked", "encoder.3.weight", "encoder.3.bias", "encoder.5.weight", "encoder.5.bias", "encoder.5.running_mean", "encoder.5.running_var", "encoder.5.num_batches_tracked", "encoder.6.weight", "encoder.6.bias", "encoder.8.weight", "encoder.8.bias", "encoder.8.running_mean", "encoder.8.running_var", "encoder.8.num_batches_tracked", "encoder.10.weight", "encoder.10.bias", "encoder.12.weight", "encoder.12.bias", "encoder.12.running_mean", "encoder.12.running_var", "encoder.12.num_batches_tracked", "encoder.13.weight", "encoder.13.bias", "encoder.15.weight", "encoder.15.bias", "encoder.15.running_mean", "encoder.15.running_var", "encoder.15.num_batches_tracked", "encoder.16.weight", "encoder.16.bias", "linear.0.weight", "linear.0.bias", "linear.2.weight", "linear.2.bias".
@MicaelJungo 回答了
您保存的权重不是来自您在此处使用的模型。确保加载正确的检查点，该检查点是在训练此特定模型时创建的。
我在 open ai gym 中创建了一个自定义环境，我在加载重量时遇到错误有人可以帮助我解决这个问题吗？我正在自定义环境中训练 TD3 网络，我已经成功训练，但在推理时我遇到了这个问题
class Actor(nn.Module):
def __init__(self, state_dim, action_dim, max_action):
super(Actor, self).__init__()
self.layer_1 = nn.Linear(state_dim, 400)
self.layer_2 = nn.Linear(400, 300)
self.layer_3 = nn.Linear(300, action_dim)
self.max_action = max_action
def forward(self, x):
x = F.relu(self.layer_1(x))
x = F.relu(self.layer_2(x))
x = self.max_action * torch.tanh(self.layer_3(x))
return x
class Critic(nn.Module):
def __init__(self, state_dim, action_dim):
super(Critic, self).__init__()
# Defining the first Critic neural network
self.layer_1 = nn.Linear(state_dim + action_dim, 400)
self.layer_2 = nn.Linear(400, 300)
self.layer_3 = nn.Linear(300, 1)
# Defining the second Critic neural network
self.layer_4 = nn.Linear(state_dim + action_dim, 400)
self.layer_5 = nn.Linear(400, 300)
self.layer_6 = nn.Linear(300, 1)
def forward(self, x, u):
xu = torch.cat([x, u], 1)
# Forward-Propagation on the first Critic Neural Network
x1 = F.relu(self.layer_1(xu))
x1 = F.relu(self.layer_2(x1))
x1 = self.layer_3(x1)
# Forward-Propagation on the second Critic Neural Network
x2 = F.relu(self.layer_4(xu))
x2 = F.relu(self.layer_5(x2))
x2 = self.layer_6(x2)
return x1, x2
def Q1(self, x, u):
xu = torch.cat([x, u], 1)
x1 = F.relu(self.layer_1(xu))
x1 = F.relu(self.layer_2(x1))
x1 = self.layer_3(x1)
return x1
# Selecting the device (CPU or GPU)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Building the whole Training Process into a class
class TD3(object):
def __init__(self, state_dim, action_dim, max_action):
self.actor = Actor(state_dim, action_dim, max_action).to(device)
self.actor_target = Actor(state_dim, action_dim, max_action).to(device)
self.actor_target.load_state_dict(self.actor.state_dict())
self.actor_optimizer = torch.optim.Adam(self.actor.parameters())
self.critic = Critic(state_dim, action_dim).to(device)
self.critic_target = Critic(state_dim, action_dim).to(device)
self.critic_target.load_state_dict(self.critic.state_dict())
self.critic_optimizer = torch.optim.Adam(self.critic.parameters())
self.max_action = max_action
def select_action(self, state):
state = torch.Tensor(state.reshape(1, -1)).to(device)
return self.actor(state).cpu().data.numpy().flatten()
def train(self, replay_buffer, iterations, batch_size=100, discount=0.99, tau=0.005, policy_noise=0.2, noise_clip=0.5, policy_freq=2):
for it in range(iterations):
# Step 4: We sample a batch of transitions (s, s’, a, r) from the memory
batch_states, batch_next_states, batch_actions, batch_rewards, batch_dones = replay_buffer.sample(batch_size)
state = torch.Tensor(batch_states).to(device)
next_state = torch.Tensor(batch_next_states).to(device)
action = torch.Tensor(batch_actions).to(device)
reward = torch.Tensor(batch_rewards).to(device)
done = torch.Tensor(batch_dones).to(device)
# Step 5: From the next state s’, the Actor target plays the next action a’
next_action = self.actor_target(next_state)
# Step 6: We add Gaussian noise to this next action a’ and we clamp it in a range of values supported by the environment
noise = torch.Tensor(batch_actions).data.normal_(0, policy_noise).to(device)
noise = noise.clamp(-noise_clip, noise_clip)
next_action = (next_action + noise).clamp(-self.max_action, self.max_action)
# Step 7: The two Critic targets take each the couple (s’, a’) as input and return two Q-values Qt1(s’,a’) and Qt2(s’,a’) as outputs
target_Q1, target_Q2 = self.critic_target(next_state, next_action)
# Step 8: We keep the minimum of these two Q-values: min(Qt1, Qt2)
target_Q = torch.min(target_Q1, target_Q2)
# Step 9: We get the final target of the two Critic models, which is: Qt = r + γ * min(Qt1, Qt2), where γ is the discount factor
target_Q = reward + ((1 - done) * discount * target_Q).detach()
# Step 10: The two Critic models take each the couple (s, a) as input and return two Q-values Q1(s,a) and Q2(s,a) as outputs
current_Q1, current_Q2 = self.critic(state, action)
# Step 11: We compute the loss coming from the two Critic models: Critic Loss = MSE_Loss(Q1(s,a), Qt) + MSE_Loss(Q2(s,a), Qt)
critic_loss = F.mse_loss(current_Q1, target_Q) + F.mse_loss(current_Q2, target_Q)
# Step 12: We backpropagate this Critic loss and update the parameters of the two Critic models with a SGD optimizer
self.critic_optimizer.zero_grad()
critic_loss.backward()
self.critic_optimizer.step()
# Step 13: Once every two iterations, we update our Actor model by performing gradient ascent on the output of the first Critic model
if it % policy_freq == 0:
actor_loss = -self.critic.Q1(state, self.actor(state)).mean()
self.actor_optimizer.zero_grad()
actor_loss.backward()
self.actor_optimizer.step()
# Step 14: Still once every two iterations, we update the weights of the Actor target by polyak averaging
for param, target_param in zip(self.critic.parameters(), self.critic_target.parameters()):
target_param.data.copy_(tau * param.data + (1 - tau) * target_param.data)
# Step 15: Still once every two iterations, we update the weights of the Critic target by polyak averaging
for param, target_param in zip(self.actor.parameters(), self.actor_target.parameters()):
target_param.data.copy_(tau * param.data + (1 - tau) * target_param.data)
# Making a save method to save a trained model
def save(self, filename, directory):
torch.save(self.actor.state_dict(), '%s/%s_actor.pth' % (directory, filename))
torch.save(self.critic.state_dict(), '%s/%s_critic.pth' % (directory, filename))
# Making a load method to load a pre-trained model
def load(self, filename, directory):
self.actor.load_state_dict(torch.load('%s/%s_actor.pth' % (directory, filename)))
self.critic.load_state_dict(torch.load('%s/%s_critic.pth' % (directory, filename)))
def evaluate_policy(policy, eval_episodes=10):
avg_reward = 0.
for _ in range(eval_episodes):
obs = env.reset()
done = False
while not done:
action = policy.select_action(np.array(obs))
obs, reward, done, _ = env.step(action)
avg_reward += reward
avg_reward /= eval_episodes
print ("---------------------------------------")
print ("Average Reward over the Evaluation Step: %f" % (avg_reward))
print ("---------------------------------------")
return avg_reward
env_name = "Pygame-v0"
seed = 0
file_name = "%s_%s_%s" % ("TD3", env_name, str(seed))
print ("---------------------------------------")
print ("Settings: %s" % (file_name))
print ("---------------------------------------")
eval_episodes = 10
save_env_vid = True
env = gym.make(env_name)
max_episode_steps = env._max_episode_steps
if save_env_vid:
env = wrappers.Monitor(env, monitor_dir, force = True)
env.reset()
env.seed(seed)
torch.manual_seed(seed)
np.random.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
max_action = float(env.action_space.high[0])
policy = TD3(state_dim, action_dim, max_action)
#policy.load(file_name, './pytorch_models/')
policy.load(file_name,"/content/gdrive/My Drive/reinforce/gym_game/pytorch_models")
_ = evaluate_policy(policy, eval_episodes=eval_episodes)
回溯： 我在为演员模型加载 state_dict 时遇到运行时错误。我搜索了 google 但找不到类似的问题。
RuntimeError: Error(s) in loading state_dict for Actor:
Missing key(s) in state_dict: "layer_1.weight", "layer_1.bias", "layer_2.weight", "layer_2.bias", "layer_3.weight", "layer_3.bias".
Unexpected key(s) in state_dict: "encoder.0.weight", "encoder.0.bias", "encoder.2.weight", "encoder.2.bias", "encoder.2.running_mean", "encoder.2.running_var", "encoder.2.num_batches_tracked", "encoder.3.weight", "encoder.3.bias", "encoder.5.weight", "encoder.5.bias", "encoder.5.running_mean", "encoder.5.running_var", "encoder.5.num_batches_tracked", "encoder.6.weight", "encoder.6.bias", "encoder.8.weight", "encoder.8.bias", "encoder.8.running_mean", "encoder.8.running_var", "encoder.8.num_batches_tracked", "encoder.10.weight", "encoder.10.bias", "encoder.12.weight", "encoder.12.bias", "encoder.12.running_mean", "encoder.12.running_var", "encoder.12.num_batches_tracked", "encoder.13.weight", "encoder.13.bias", "encoder.15.weight", "encoder.15.bias", "encoder.15.running_mean", "encoder.15.running_var", "encoder.15.num_batches_tracked", "encoder.16.weight", "encoder.16.bias", "linear.0.weight", "linear.0.bias", "linear.2.weight", "linear.2.bias".
@MicaelJungo 回答了
您保存的权重不是来自您在此处使用的模型。确保加载正确的检查点，该检查点是在训练此特定模型时创建的。