Policy distillation: Example 1

参考：

GitHub - CUN-bjy/policy-distillation-baselines: Pytorch Implementation of Policy Distillation for control, which has well-trained teachers via stable_baselines3.

 

配置conda环境：

python==3.7
gym==0.19.0
tensorboardX
torch
stable-baselines3[extra,tests,docs]==1.1.0
pybullet==2.7.8
optuna
pyyaml>=5.1
sb3-contrib==1.0.0

测试教师模型：

python policy_distillation.py --model teacher --algo td3 --env AntBulletEnv-v0

知识蒸馏：

先创建文件夹：

mkdir distilled-agents

 训练：

python policy_distillation.py --algo td3 --env AntBulletEnv-v0

测试学生：

python playground.py --mode student -p /home/blamlight/Documents/Github/policy-distillation-baselines/distilled-agents/AntBulletEnv-v0_td3_1710318896.1154015/student_10000_3260.12.pkl