06/28 인공지능 펭귄만들기

1. 환경 변수 확인

• 파이썬 path가 잡혀있는지 확인
• 3.62 확인

cmd

python -V

where python

 

2. ml agent 다운로드

workspace/unity에 저장

https://github.com/Unity-Technologies/ml-agents

Unity-Technologies/ml-agents

 

ml agent 세팅

https://cafe.naver.com/gameprogramming7

종로 더조은 게임 개발자 과정 3기 : 네이버 카페

D:\workspace\unity\ml-agents-release_17\ml-agents-release_17\com.unity.ml-agents

package.json

 

 

3. 튜토리얼 진행

Asset/Meshes

펭귄 에셋 다운 및 튜토리얼

https://www.immersivelimit.com/tutorials/reinforcement-learning-penguins-part-1-unity-ml-agents

Reinforcement Learning Penguins (Part 1/4) | Unity ML-Agents — Immersive Limit

or

카페

https://cafe.naver.com/gameprogramming7/413

인공지능

 

4. 훈련시작

Penguin.yaml

behaviors:
    Penguin:
        trainer_type: ppo
        hyperparameters:
            batch_size: 128
            buffer_size: 2048
            learning_rate: 0.0003
            beta: 0.01
            epsilon: 0.2
            lambd: 0.95
            num_epoch: 3
            learning_rate_schedule: linear
        network_settings:
            normalize: false
            hidden_units: 256
            num_layers: 2
            vis_encode_type: simple
        reward_signals:
            extrinsic:
                gamma: 0.99
                strength: 1.0
        keep_checkpoints: 5
        max_steps: 1000000
        time_horizon: 128
        summary_freq: 5000
        threaded: true

 

mlagents-learn [Penguin.yaml위치 지정] --run-id Penguin_01

mlagents-learn config/ppo/Penguin.yaml --run-id Penguin_01

 

mlagents가 작동 안할 시

pip3 install torch~=1.7.1 -f https://download.pytorch.org/whl/torch_stable.html

안되면

python -m pip install mlagents==0.26.0

 

중지

mlagents 가 실행중인 cmd에서 컨트롤 + c

 

재시작 시

mlagents-learn [Penguin.yaml위치 지정] --run-id Penguin_01 --resume

 

 

훈련된 파일 

D:\workspace\unity\ml-agents-release_17\ml-agents-release_17\results\Penguin_01\Penguin.onnx

model에 Penguin.onnx를 어사인하면 훈련된 결과를 볼 수 있음