Competition_3v3snakes

Environment

Check details in Jidi Competition RLChina2021智能体竞赛

Dependency

You need to create competition environment.

conda create -n snake3v3 python=3.7.5

conda activate snake3v3

pip install -r requirements.txt

How to train rl-agent

python rl_trainer/main.py

By default-parameters, the total reward of training is shown below.

You can edit different parameters, for example

python rl_trainer/main.py --algo "bicnet" --epsilon 0.8

Baseline performance:

You can locally evaluation your model.

python evaluation_local.py --my_ai rl --opponent random

How to test submission

You can locally test your submission. At Jidi platform, we evaluate your submission as same as run_log.py

Once you run this file, you can locally check battle logs in the folder named "logs".

For example,

python run_log.py --my_ai "random" --opponent "rl"

Ready to submit

1. Random policy --> agent/random/submission.py
2. RL policy --> all files in agent/rl/

Watch reply locally

1. Open reply/reply.html in any browser.
2. Load a log.
3. Reply and watch ^0^.