interpreting-rewards/prepare at master · HumanCompatibleAI/interpreting-rewards · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
#!/usr/bin/env python3

# --------------------------------------------------------------------
# Prepare the rl-baselines3-zoo submodule for use with mazelab.
# --------------------------------------------------------------------

hyperparams = r"""
CoinFlipGoalWithGroundTruth:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 45000
  normalize: false
  learning_rate: lin_1.0e-4

CoinFlipGoalWithCoinFlipGoalRewardModel:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 45000
  normalize: false
  learning_rate: lin_1.0e-4
  env_wrapper:
    - interp.common.wrappers.WrapWithRewardModel:
        model_path: "../reward-models/EmptyMaze-10x10-CoinFlipGoal-v3-reward_model.pt"

TwoGoalsWithGroundTruth:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 45000
  normalize: false
  learning_rate: lin_1.0e-4

TwoGoalsWithCoinFlipGoalRewardModel:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 45000
  normalize: false
  learning_rate: lin_1.0e-4
  env_wrapper:
    - interp.common.wrappers.WrapWithRewardModel:
        model_path: "../reward-models/EmptyMaze-10x10-CoinFlipGoal-v3-reward_model.pt"


EmptyMaze-10x10-FixedGoal-v3:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 30000
  normalize: false
  learning_rate: lin_1.0e-4

EmptyMaze-10x10-FixedGoal-v3:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 30000
  normalize: false
  learning_rate: lin_1.0e-4

EmptyMaze-10x10-CoinFlipGoal-v3:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 45000
  normalize: false
  learning_rate: lin_1.0e-4

EmptyMaze-10x10-RandomGoal-v3:
  policy: 'MlpPolicy'
  gamma: 0.97
  n_timesteps: !!float 250000
  normalize: false
  learning_rate: lin_1.0e-4

"""

if __name__ == '__main__':
    with open('rl-baselines3-zoo/hyperparams/ppo.yml', 'r') as f:
        original = f.read()
        new = hyperparams + original
    with open('rl-baselines3-zoo/hyperparams/ppo.yml', 'w') as f:
        f.write(new)