ppo Agent playing Pyramids

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

Watch the Agent play

You can watch the agent playing directly in your browser

Go to https://huggingface.co./spaces/unity/ML-Agents-Pyramids Step 1: Find the model_id: Francesco-A/ppo-Pyramids-v1 Step 2: Select the .nn /.onnx file Click on Watch the agent play

Resume the training

mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume

Training hyperparameters

behaviors:
  Pyramids:
    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: 512
      num_layers: 2
      vis_encode_type: simple
    reward_signals:
      extrinsic:
        gamma: 0.99
        strength: 1.0
      rnd:
        gamma: 0.99
        strength: 0.01
        network_settings:
          hidden_units: 64
          num_layers: 3
        learning_rate: 0.0001
    keep_checkpoints: 5
    max_steps: 1000000
    time_horizon: 128
    summary_freq: 30000

Training details

Step	Time Elapsed	Mean Reward	Std of Reward	Status
30000	59.481 s	-1.000	0.000	Training
60000	118.648 s	-0.798	0.661	Training
90000	180.684 s	-0.701	0.808	Training
120000	240.734 s	-0.931	0.373	Training
150000	300.978 s	-0.851	0.588	Training
180000	360.137 s	-0.934	0.361	Training
210000	424.326 s	-1.000	0.000	Training
240000	484.774 s	-0.849	0.595	Training
270000	546.089 s	-0.377	1.029	Training
300000	614.797 s	-0.735	0.689	Training
330000	684.241 s	-0.926	0.405	Training
360000	745.790 s	-0.819	0.676	Training
390000	812.573 s	-0.715	0.755	Training
420000	877.836 s	-0.781	0.683	Training
450000	944.423 s	-0.220	1.114	Training
480000	1010.918 s	-0.484	0.962	Training
510000	1074.058 s	-0.003	1.162	Training
540000	1138.848 s	-0.021	1.222	Training
570000	1204.326 s	0.384	1.231	Training
600000	1276.488 s	0.690	1.174	Training
630000	1345.297 s	0.943	1.058	Training
660000	1412.791 s	1.014	1.043	Training
690000	1482.712 s	0.927	1.054	Training
720000	1548.726 s	0.900	1.128	Training
750000	1618.284 s	1.379	0.701	Training
780000	1692.080 s	1.567	0.359	Training
810000	1762.159 s	1.475	0.567	Training
840000	1832.166 s	1.438	0.648	Training
870000	1907.191 s	1.534	0.536	Training
900000	1977.521 s	1.552	0.478	Training
930000	2051.259 s	1.458	0.633	Training
960000	2126.498 s	1.545	0.586	Training
990000	2198.591 s	1.565	0.591	Training