ppo Agent playing SnowballTarget

This is a trained model of a ppo agent playing SnowballTarget using the Unity ML-Agents Library.

Usage (with ML-Agents)

The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub:

Resume the training

mlagents-learn ./config/ppo/SnowballTarget.yaml --run-id="SnowballTarget-v1" --resume

Training hyperparameters

behaviors:
SnowballTarget:
  trainer_type: ppo
  summary_freq: 10000
  keep_checkpoints: 5
  checkpoint_interval: 50000
  max_steps: 900000
  time_horizon: 128
  threaded: true
  hyperparameters:
    learning_rate: 0.0001
    learning_rate_schedule: linear
    batch_size: 128
    buffer_size: 4096
    beta: 0.005
    epsilon: 0.2
    lambd: 0.95
    num_epoch: 5
  network_settings:
    normalize: false
    hidden_units: 256
    num_layers: 3
    vis_encode_type: simple
  reward_signals:
    extrinsic:
      gamma: 0.99
      strength: 1.0

Watch your Agent play

You can watch your agent playing directly in your browser:.

Go to https://huggingface.co/spaces/unity/ML-Agents-SnowballTarget
Step 1: Write your model_id: kinkpunk/PPO-SnowballTarget
Step 2: Select your .nn /.onnx file
Click on Watch the agent play 👀