ML-Agents stopped before arrive to Max steps.

Unity Engine
, ,
1

Hello.

I am trying to use ML-Agents.
Now, ML-Agents learnning stopped before arrive to Max steps.
(Stop at 500000 steps)

config file is follows:

behaviors:
  BalanceGame:
    trainer_type: ppo
    hyperparameters:
      batch_size: 512
      buffer_size: 65536
      learning_rate: 3e-5
      beta: 0.001
      epsilon: 0.2
      lambd: 0.99
      num_epoch: 8
      learning_rate_schedule: linear
    network_settings:
      normalize: true
      hidden_units: 256
      num_layers: 2
      vis_encode_type: simple
    reward_signals:
      extrinsic:
        gamma: 0.99
        strength: 1.0
    keep_checkpoints: 50000
    max_steps: 1.0e7
    time_horizon: 64
    summary_freq: 10000
    threaded: true

My environments are follows:
Unity: 2020.3.17f1
ML-Agents:0.17.0

Thanks for your help.

2

500000 steps is the default value. Are you sure you have the correct behaviour name or you are using this yaml file?

1 Like
3

It salved.
I missed behavior name in yaml file.
Thank you!

1 Like