Unity 机器学习代理工具包 (ML-Agents) 是一个开源项目,它使游戏和模拟能够作为训练智能代理的环境。
您最多选择25个主题 主题必须以中文或者字母或数字开头,可以包含连字符 (-),并且长度不得超过35个字符
 
 
 
 
 

27 行
1.1 KiB

behaviors:
WalkerStatic:
behavioral_cloning: null
checkpoint_interval: 500000
hyperparameters: {batch_size: 2048, beta: 0.005, buffer_size: 20480, epsilon: 0.2,
lambd: 0.95, learning_rate: 0.0003, learning_rate_schedule: linear, num_epoch: 3}
init_path: null
keep_checkpoints: 5
max_steps: 20000000
network_settings: {hidden_units: 512, memory: null, normalize: true, num_layers: 3,
vis_encode_type: simple}
reward_signals:
extrinsic: {gamma: 0.995, strength: 1.0}
self_play: null
summary_freq: 30000
threaded: true
time_horizon: 1000
trainer_type: ppo
checkpoint_settings: {force: false, inference: false, initialize_from: null, load_model: false,
resume: false, run_id: wst-ppo, train_model: true}
curriculum: null
debug: false
engine_settings: {capture_frame_rate: 60, height: 84, no_graphics: true, quality_level: 5,
target_frame_rate: -1, time_scale: 20, width: 84}
env_settings: {base_port: 5005, env_args: null, env_path: /unity-volume/executable,
num_envs: 8, seed: -1}
parameter_randomization: null