kengz/SLM-Lab

View on GitHub
slm_lab/spec/experimental/misc/mountain_car.json

Summary

Maintainability
Test Coverage
{
  "a2c_gae_mlp_separate_mountain_car": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.999,
        "lam": 0.97,
        "num_step_returns": null,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 1000,
          "end_step": 5000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 1.0,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "MLPNet",
        "shared": false,
        "hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "clip_grad_val": null,
        "use_same_optim": false,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 2600,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "lam__uniform": [0.9, 0.99]
        },
        "net": {
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[100], [200], [400], [200, 100]],
          "actor_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
          "critic_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          }
        }
      }]
    }
  },
  "a2c_gae_rnn_separate_mountain_car": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.999,
        "lam": 0.97,
        "num_step_returns": null,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 1000,
          "end_step": 5000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 1.0,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "RecurrentNet",
        "shared": false,
        "cell_type": "GRU",
        "fc_hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "rnn_hidden_size": 64,
        "rnn_num_layers": 1,
        "seq_len": 4,
        "clip_grad_val": null,
        "use_same_optim": false,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 2600,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "lam__uniform": [0.9, 0.99],
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[], [100], [200]],
          "rnn_hidden_size__choice": [32, 64, 128],
          "seq_len__choice": [2, 3, 4, 5, 6],
          "actor_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
          "critic_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          }
        }
      }]
    }
  },
  "a2c_nstep_mlp_separate_mountain_car": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.999,
        "lam": null,
        "num_step_returns": 1,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 1000,
          "end_step": 5000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 1.0,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "MLPNet",
        "shared": false,
        "hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "clip_grad_val": null,
        "use_same_optim": false,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 2600,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "num_step_returns__choice": [2, 4, 8]
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[100], [200], [400], [200, 100]],
          "actor_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
          "critic_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          }
        }
      }]
    }
  },
  "a2c_nstep_rnn_separate_mountain_car": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.999,
        "lam": null,
        "num_step_returns": 1,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 1000,
          "end_step": 5000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 1.0,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "RecurrentNet",
        "shared": false,
        "cell_type": "GRU",
        "fc_hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "rnn_hidden_size": 64,
        "rnn_num_layers": 1,
        "seq_len": 4,
        "clip_grad_val": null,
        "use_same_optim": false,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 2600,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "lam__uniform": [0.9, 0.99],
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[], [100], [200]],
          "rnn_hidden_size__choice": [32, 64, 128],
          "seq_len__choice": [2, 3, 4, 5, 6],
          "actor_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
          "critic_optim_spec": {
            "lr__uniform": [0.001, 0.01]
          }
        }
      }]
    }
  },
  "dqn_boltzmann_mountain_car": {
    "agent": [{
      "name": "DQN",
      "algorithm": {
        "name": "DQN",
        "action_pdtype": "Categorical",
        "action_policy": "boltzmann",
        "explore_var_spec": {
          "name": "linear_decay",
          "start_val": 5.0,
          "end_val": 1.0,
          "start_step": 0,
          "end_step": 80000,
        },
        "gamma": 0.999,
        "training_batch_iter": 3,
        "training_iter": 4,
        "training_frequency": 4,
        "training_start_step": 32
      },
      "memory": {
        "name": "Replay",
        "batch_size": 32,
        "max_size": 100000,
        "use_cer": false
      },
      "net": {
        "type": "MLPNet",
        "hid_layers": [200, 100],
        "hid_layers_activation": "relu",
        "clip_grad_val": null,
        "loss_spec": {
          "name": "MSELoss"
        },
        "optim_spec": {
          "name": "Adam",
          "lr": 0.006
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "update_type": "polyak",
        "update_frequency": 200,
        "polyak_coef": 0.8,
        "gpu": false,
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 1400,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "explore_var_spec": {
            "end_step__choice": [20000, 40000, 80000]
          }
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "polyak_coef__uniform": [0.8, 1.0],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[100], [200], [400], [200, 100]],
          "optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
        }
      }]
    }
  },
  "drqn_boltzmann_mountain_car": {
    "agent": [{
      "name": "DQN",
      "algorithm": {
        "name": "DQN",
        "action_pdtype": "Categorical",
        "action_policy": "boltzmann",
        "explore_var_spec": {
          "name": "linear_decay",
          "start_val": 5.0,
          "end_val": 1.0,
          "start_step": 0,
          "end_step": 40000,
        },
        "gamma": 0.999,
        "training_batch_iter": 3,
        "training_iter": 4,
        "training_frequency": 4,
        "training_start_step": 32
      },
      "memory": {
        "name": "Replay",
        "batch_size": 32,
        "max_size": 100000,
        "use_cer": false
      },
      "net": {
        "type": "RecurrentNet",
        "cell_type": "GRU",
        "fc_hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "rnn_hidden_size": 64,
        "rnn_num_layers": 1,
        "seq_len": 4,
        "clip_grad_val": null,
        "loss_spec": {
          "name": "MSELoss"
        },
        "optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "update_type": "polyak",
        "update_frequency": 200,
        "polyak_coef": 0,
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 1400,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "explore_var_spec": {
            "end_step__choice": [20000, 40000, 80000]
          }
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "polyak_coef__uniform": [0.8, 1.0],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[], [100], [200]],
          "rnn_hidden_size__choice": [32, 64, 128],
          "seq_len__choice": [2, 3, 4, 5, 6],
          "optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
        }
      }]
    }
  },
  "ddqn_boltzmann_mountain_car": {
    "agent": [{
      "name": "DoubleDQN",
      "algorithm": {
        "name": "DoubleDQN",
        "action_pdtype": "Categorical",
        "action_policy": "boltzmann",
        "explore_var_spec": {
          "name": "linear_decay",
          "start_val": 5.0,
          "end_val": 1.0,
          "start_step": 0,
          "end_step": 40000,
        },
        "gamma": 0.999,
        "training_batch_iter": 4,
        "training_iter": 4,
        "training_frequency": 4,
        "training_start_step": 32
      },
      "memory": {
        "name": "Replay",
        "batch_size": 32,
        "max_size": 100000,
        "use_cer": false
      },
      "net": {
        "type": "MLPNet",
        "hid_layers": [200, 100],
        "hid_layers_activation": "relu",
        "clip_grad_val": null,
        "loss_spec": {
          "name": "MSELoss"
        },
        "optim_spec": {
          "name": "Adam",
          "lr": 0.004
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "update_type": "polyak",
        "update_frequency": 200,
        "polyak_coef": 0.8,
        "gpu": false,
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 1400,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "explore_var_spec": {
            "end_step__choice": [20000, 40000, 80000]
          }
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "polyak_coef__uniform": [0.8, 1.0],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[100], [200], [400], [200, 100]],
          "optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
        }
      }]
    }
  },
  "ddrqn_boltzmann_mountain_car": {
    "agent": [{
      "name": "DoubleDQN",
      "algorithm": {
        "name": "DoubleDQN",
        "action_pdtype": "Categorical",
        "action_policy": "boltzmann",
        "explore_var_spec": {
          "name": "linear_decay",
          "start_val": 5.0,
          "end_val": 1.0,
          "start_step": 0,
          "end_step": 40000,
        },
        "gamma": 0.999,
        "training_batch_iter": 4,
        "training_iter": 4,
        "training_frequency": 4,
        "training_start_step": 32
      },
      "memory": {
        "name": "Replay",
        "batch_size": 32,
        "max_size": 100000,
        "use_cer": false
      },
      "net": {
        "type": "RecurrentNet",
        "cell_type": "GRU",
        "fc_hid_layers": [200],
        "hid_layers_activation": "sigmoid",
        "rnn_hidden_size": 64,
        "rnn_num_layers": 1,
        "seq_len": 4,
        "clip_grad_val": null,
        "loss_spec": {
          "name": "MSELoss"
        },
        "optim_spec": {
          "name": "Adam",
          "lr": 0.01
        },
        "lr_scheduler_spec": {
          "name": "StepLR",
          "step_size": 160000,
          "gamma": 0.9,
        },
        "update_type": "polyak",
        "update_frequency": 200,
        "polyak_coef": 0,
        "gpu": false
      }
    }],
    "env": [{
      "name": "MountainCar-v0",
      "max_t": null,
      "max_frame": 1400,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 200,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "explore_var_spec": {
            "end_step__choice": [20000, 40000, 80000]
          }
        },
        "net": {
          "lr_decay_frequency__choice": [5000, 10000, 50000, 10000],
          "polyak_coef__uniform": [0.8, 1.0],
          "hid_layers_activation__choice": ["relu", "selu", "sigmoid"],
          "hid_layers__choice": [[], [100], [200]],
          "rnn_hidden_size__choice": [32, 64, 128],
          "seq_len__choice": [2, 3, 4, 5, 6],
          "optim_spec": {
            "lr__uniform": [0.001, 0.01]
          },
        }
      }]
    }
  },
}