英文:
Create Custom Policy using stable baseline3
问题
我正在尝试创建一个自定义的LSTM策略。似乎缺少BasePolicy。我们如何创建一个自定义的LSTM策略以传递给PPO或A2C算法。此外，如果不行，是否可以修改当前设置中LSTM层，以帮助自定义我的结果。
import gym
import torch.nn as nn
from sb3_contrib.common.policies import BasePolicy
from sb3_contrib.ppo_recurrent import RecurrentPPO
class CustomPolicy(BasePolicy):
def __init__(self, *args, **kwargs):
super(CustomPolicy, self).__init__(*args, **kwargs)
def make_lstm_layer(self, n_lstm_layers: int) -> nn.Module:
# 在这里创建您的自定义LSTM层
# 例如：
lstm_layer = nn.LSTM(input_size=self.features_dim, hidden_size=64, num_layers=n_lstm_layers)
return lstm_layer
env_name = "CartPole-v1"
env = gym.make(env_name)
model = RecurrentPPO(CustomPolicy, env, verbose=1)
请指导。
我尝试安装这些包，但无法弄清楚如何创建自定义策略或修改他们提供的策略。
英文:
I am trying to create a custom lstm policy. It seems that BasePolicy is missing. How can we create a custom LSTM policy to pass to PPO or A2C algorithm. Also, if not, can modify the layer of lstm in the current setting that will help in customizing my results.
import gym
import torch.nn as nn
from sb3_contrib.common.policies import BasePolicy
from sb3_contrib.ppo_recurrent import RecurrentPPO
class CustomPolicy(BasePolicy):
def __init__(self, *args, **kwargs):
super(CustomPolicy, self).__init__(*args, **kwargs)
def make_lstm_layer(self, n_lstm_layers: int) -> nn.Module:
# Create your custom LSTM layer here
# For example:
lstm_layer = nn.LSTM(input_size=self.features_dim, hidden_size=64, num_layers=n_lstm_layers)
return lstm_layer
env_name = "CartPole-v1"
env = gym.make(env_name)
model = RecurrentPPO(CustomPolicy, env, verbose=1)
Please guide.
I tried installing the packages but was unable to figure out how to create the custom policies or modify the policies they have provided
答案1
得分: 1
LSTM-policies不受sb3默认支持，但它们受到来自sb3-contrib的RecurrentPPO支持。
英文:
LSTM-policies are not supported by sb3 out-of-the-box, but they are supported by RecurrentPPO from sb3-contrib.
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论