How to save and use the pre-trained DQN agent in the reinforcement learning tool box

Question

Shuhei Inada 2021-1-9

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/712518-how-to-save-and-use-the-pre-trained-dqn-agent-in-the-reinforcement-learning-tool-box

评论： Maha Mosalam 2022-3-23

I would like to use the DQN agent pre-trained in the following neural network to solve other problems. How can I save the pre-trained DQN agent. Also, how can I use this trained agent to solve other problems.

lgraph = layerGraph();

tempLayers = [

imageInputLayer([1 1 1],"Name","Block Actions","Normalization","none")

fullyConnectedLayer(50,"Name","fc_2")

reluLayer("Name","relu_2")];

lgraph = addLayers(lgraph,tempLayers);

tempLayers = [

imageInputLayer([20 1 1],"Name","Block Observations","Normalization","none")

scalingLayer("Name","scaling","Scale",12)

fullyConnectedLayer(50,"Name","fc_1")

reluLayer("Name","relu_1")];

lgraph = addLayers(lgraph,tempLayers);

tempLayers = [

additionLayer(2,"Name","addition")

fullyConnectedLayer(10,"Name","fc_3")

reluLayer("Name","relu_3")

fullyConnectedLayer(1,"Name","stateValue")];

lgraph = addLayers(lgraph,tempLayers);

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Emmanouil Tzorakoleftherakis 2021-1-16

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/712518-how-to-save-and-use-the-pre-trained-dqn-agent-in-the-reinforcement-learning-tool-box#answer_599520

在 MATLAB Online 中打开

Hello,

Take a look at this example, and specifically the code snippet below:

if doTraining    
    % Train the agent.
    trainingStats = train(agent,env,trainOpts);
    % Save the agent
    save myagent.mat agent
else
    % Load the pretrained agent for the example.
    load('MATLABCartpoleDQNMulti.mat','agent')
end

3 个评论
显示 1更早的评论隐藏 1更早的评论

Maha Mosalam 2022-3-22

So when I use the pretrained agent after that to get actions, I should defined the whole algorithm , I mean(the environment action space , actor and critic layers , expect the trainig line , just

use load(myagent.mat, agent)

Maha Mosalam 2022-3-23

Warning: Cannot load an object of class 'rlDLNetworkModel':

Its class cannot be found.

Warning: Cannot load an object of class 'rlADAMOptimizer':

Its class cannot be found.

Warning: While loading an object of class 'rl.representation.rlQValueRepresentation':

Undefined function 'getLoss' for input arguments of type 'double'.

Warning: Cannot load an object of class 'rlDLNetworkModel':

Its class cannot be found.

.....

I got these mesege when just

load(myagent.mat, agent)

then get action from that, not redefined the actor and critic, and environment

请先登录，再进行评论。

How to save and use the pre-trained DQN agent in the reinforcement learning tool box

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

回答（1 个）

3 个评论
显示 1更早的评论隐藏 1更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

How to save and use the pre-trained DQN agent in the reinforcement learning tool box

0 个评论 显示 -2更早的评论隐藏 -2更早的评论

回答（1 个）

3 个评论 显示 1更早的评论隐藏 1更早的评论

另请参阅

类别

标签

产品

版本

Community Treasure Hunt

WeChat

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

3 个评论
显示 1更早的评论隐藏 1更早的评论