Skip to content

/ deeplearning4j

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

The Cartpole example can't converge #8727

Open
hanbaoan123 opened this issue Feb 24, 2020 · 4 comments
Open

The Cartpole example can't converge #8727

hanbaoan123 opened this issue Feb 24, 2020 · 4 comments
Assignees
@saudet @RobAltena
Labels
Documentation RL4J

Comments

@hanbaoan123
Copy link

@hanbaoan123 hanbaoan123 commented Feb 24, 2020

Issue Description

When I run the example cartpole with the default parameters, it can not converge to the max reward 200, I wonder what went wrong.
360截图20200224095510956

Version Information

Please indicate relevant versions, including, if relevant:

  • Deeplearning4j version beta6
  • Platform information (OS, etc) windows 10
  • CUDA version, if used CUDA 10.0
  • NVIDIA driver version, if in use 431.40

Additional Information

Where applicable, please also provide:

  • Full log or exception stack trace (ideally in a Gist: gist.github.com)
  • pom.xml file or similar (also in a Gist)
@saudet
Copy link
Member

@saudet saudet commented Feb 24, 2020

@RobAltena
@saudet saudet self-assigned this Feb 24, 2020
@RobAltena
Copy link
Contributor

@RobAltena RobAltena commented Feb 24, 2020

There is an updated cartpole example here: https://github.com/RobAltena/cartpole

But it uses the daily snapshots for now.
@hanbaoan123
Copy link
Author

@hanbaoan123 hanbaoan123 commented Feb 24, 2020

There is an updated cartpole example here: https://github.com/RobAltena/cartpole

But it uses the daily snapshots for now.

So what's going on? Are the parameters inappropriate？
@RobAltena
Copy link
Contributor

@RobAltena RobAltena commented Feb 24, 2020

I do not know what exactly was wrong with the old example. But the new one uses an entirely different neural network. I has more nodes and we use a different optimizer.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@saudet saudet

@RobAltena RobAltena

Labels
Documentation RL4J
Projects
None yet
Milestone
No milestone
Linked pull requests

Successfully merging a pull request may close this issue.

None yet
3 participants
@saudet @RobAltena @hanbaoan123
You can’t perform that action at this time.