need more explanations for arguments

[zinwalin]

usage: run.py [-h] [-s {greedy,hamilton,dqn}]
              [-m {normal,bcmk,train_dqn,train_dqn_gui}]

Run snake game agent.

optional arguments:
  -h, --help            show this help message and exit
  -s {greedy,hamilton,dqn}
                        name of the solver to direct the snake (default:
                        hamilton)
  -m {normal,bcmk,train_dqn,train_dqn_gui}
                        game mode (default: normal)

how long it will take to run with dqn solver?

python run.py -s dqn -m train_dqn_gui

[chuyangliu]

It depends on the capacity of the machine (e.g., CPU, GPU, Memory) you use and how many steps you want to train the model. Training with GUI is just to give you a visual image of how the snake learns the steps. It's bad for training speed since rendering the GUI takes a lot of time. I would recommend using train_dqn instead of train_dqn_gui if you want to train the model faster.

If you want details of how to train the model or have any other questions, please feel free to follow up. 👍

[zinwalin]

It depends on the capacity of the machine (e.g., CPU, GPU, Memory) you use and how many steps you want to train the model. Training with GUI is just to give you a visual image of how the snake learns the steps. It's bad for training speed since rendering the GUI takes a lot of time. I would recommend using train_dqn instead of train_dqn_gui if you want to train the model faster.

If you want details of how to train the model or have any other questions, please feel free to follow up. 👍

it's a lot faster with train_dqn, thanks.

[zinwalin]

How can I check performance after training with train_dqn mode? I mean how can run in dqn mode with trained models?

[chuyangliu]

34