Hi all ,
I am planning to use rllib to train a custom external enviroment. Firstly i run the example Cartpole_server.py script in colab, and the Cartpole_client.py script in my local laptop. Then i connect to the colab public ip via ssh and remap the localhost port to my local laptop. But the client could not recive the correct data but raise an error.
I am so grateful if any friends could have some advice about how to depoly a external envs and train it in the cloud, thank you so much!!
(1)the output in colab of the server script:
(pid=1252) 2021-11-15 07:24:08,211
INFO policy_server_input.py:190 – Sending worker creation args to client.
(pid=1252) 127.0.0.1 - - [15/Nov/2021 07:24:08] “POST / HTTP/1.1” 200 -
(2)error occoured in the client script in local laptop.
Hi @Slientes ,
in my opinion using Google Colab is more a hassle than making things easier for you. I run my experiments in the Google Cloud by using a ray cluster on some servers from Compute Engine.
See this topic for more infos on how to do it.
Hope this helps
1 Like
Hi @Lars_Simon_Zehnder ,
Thank you so much for your advice. I think this is quit helpful for me, and i will make an attempt to arrange my project follow your gudience.
Good luck!!
1 Like