Skip to content

Training-Process.md

Latest commit

 

History

History
33 lines (27 loc) · 1.28 KB

Training-Process.md

File metadata and controls

33 lines (27 loc) · 1.28 KB
Raw

Training Process

Reinforcement Learning

Using Proximal Policy Optimzation (PPO)

  • Locates learn_rl.py
  • Modify the parameters (more info)
  • Run learn_rl.py

General Machine Learning

  • Locates learn_ml.py
  • Modify the parameters (more info)
  • Implement your own decision algorithm in def decide(brain_info: BrainInfo) function
  • Run learn_ml.py

OpenAI Gym Compitable

  • Sample Code: learn_gym.py
from gym_unity.envs import UnityEnv

env = UnityEnv(environment_filename, worker_id=0, use_visual, multiagent, env_config,camera_res_overwrite)
  • Limitation: By default the first visual observation is provided as the observation, if present. Otherwise vector observations are provided.
    All BrainInfo output from the environment can still be accessed from the info provided by env.step(action)

Inference

  • Set load_model = True Load the pre-train model
  • Set train_model = False Don't run any learning algorithm
  • Set fast_simulation = False Enable inference mode, allow you to use WASD-controled Observe Camera
  • Run learn_rl.py

Runing the Pre-train model