You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
Thanks for your Awesome work. Very interesting
Can you please explain in short how do you run this entire program? I am quite new to Reinforcement learning. I can see there are 2 set of files: One Specifically for the environment and the other for the RL algorithm. I use Ubuntu 18.04 so I feel there will not be any additional requirements to be installed.
Waiting to hear from you.
Hello Karthik!
Did you figure out how to run it?
I ran it but I am getting errors about registering the environment.
About your question about how to run:
You just need to run python3 main.py inside JSS folder. About environment, since a package is already mentioned in the requirements.txt to install for the environment so there is nothing you need to do for that.
This project is a simple training approach using PPO to train an agent on the Job-Shop Scheduling Environment we have developed.
If you run the main file while having previously set up the configuration dictionary correctly (i.e., the path to the file containing the instance you want to solve and other hyperparameters). It will train an agent to solve the given instance and report the best solution found using Weights and Biases.
Activity
smita-09 commentedon Mar 4, 2022
Hello Karthik!
Did you figure out how to run it?
I ran it but I am getting errors about registering the environment.
About your question about how to run:
You just need to run
python3 main.pyinside JSS folder. About environment, since a package is already mentioned in the requirements.txt to install for the environment so there is nothing you need to do for that.ingambe commentedon Mar 5, 2022
Hi @karthik7194,
First of all, sorry for the late answer :(
This project is a simple training approach using PPO to train an agent on the Job-Shop Scheduling Environment we have developed.
If you run the main file while having previously set up the configuration dictionary correctly (i.e., the path to the file containing the instance you want to solve and other hyperparameters). It will train an agent to solve the given instance and report the best solution found using Weights and Biases.
The GIF generation in itself is nothing very specific to RL. The full code to generate a GIF is available here: https://github.com/prosysscience/JSSEnv/blob/master/tests/test_rendering.py
You can replace the file with your given instance, provide the agent's solution, and generate a GIF.
I hope it helps :)