kengz/SLM-Lab

{
  "a2c_gae_separate_pendulum": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.95,
        "lam": 0.97,
        "num_step_returns": null,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 60000,
          "end_step": 120000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 0.01,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "MLPNet",
        "shared": false,
        "hid_layers": [400, 200],
        "hid_layers_activation": "relu",
        "clip_grad_val": null,
        "use_same_optim": true,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.005
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.005
        },
        "lr_scheduler_spec": {
          "name": "MultiStepLR",
          "milestones": [80000],
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "Pendulum-v0",
      "max_t": null,
      "max_frame": 500000,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 190,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "lam__uniform": [0.95, 0.99],
          "entropy_coef_spec": {
            "start_val__uniform": [0.001, 0.05],
            "end_val__uniform": [0.0001, 0.001],
            "start_step__choice": [100000, 125000, 150000],
            "end_step__choice": [150000, 175000, 200000, 250000],
          },
        },
        "net": {
          "hid_layers__choice": [[400, 200], [256, 128]],
          "actor_optim_spec": {
           "lr__choice": [0.0001, 0.0005, 0.001, 0.005, 0.01]
         },
        }
      }]
    }
  },
  "a2c_nstep_separate_pendulum": {
    "agent": [{
      "name": "A2C",
      "algorithm": {
        "name": "ActorCritic",
        "action_pdtype": "default",
        "action_policy": "default",
        "explore_var_spec": null,
        "gamma": 0.95,
        "lam": null,
        "num_step_returns": 1,
        "entropy_coef_spec": {
          "name": "linear_decay",
          "start_val": 0.01,
          "end_val": 0.001,
          "start_step": 60000,
          "end_step": 120000,
        },
        "policy_loss_coef": 1.0,
        "val_loss_coef": 0.01,
        "training_frequency": 1,
      },
      "memory": {
        "name": "OnPolicyReplay"
      },
      "net": {
        "type": "MLPNet",
        "shared": false,
        "hid_layers": [400, 200],
        "hid_layers_activation": "relu",
        "clip_grad_val": null,
        "use_same_optim": true,
        "actor_optim_spec": {
          "name": "Adam",
          "lr": 0.005
        },
        "critic_optim_spec": {
          "name": "Adam",
          "lr": 0.005
        },
        "lr_scheduler_spec": {
          "name": "MultiStepLR",
          "milestones": [80000],
          "gamma": 0.9,
        },
        "gpu": false
      }
    }],
    "env": [{
      "name": "Pendulum-v0",
      "max_t": null,
      "max_frame": 500000,
    }],
    "body": {
      "product": "outer",
      "num": 1
    },
    "meta": {
      "distributed": false,
      "eval_frequency": 1000,
      "max_session": 4,
      "max_trial": 190,
      "search": "RandomSearch",
    },
    "search": {
      "agent": [{
        "algorithm": {
          "training_frequency__choice": [2, 4, 8],
          "num_step_returns__choice": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
          "entropy_coef_spec": {
            "start_val__uniform": [0.001, 0.05],
            "end_val__uniform": [0.0001, 0.001],
            "start_step__choice": [100000, 125000, 150000],
            "end_step__choice": [150000, 175000, 200000, 250000],
          },
        },
        "net": {
          "hid_layers__choice": [[400, 200], [256, 128]],
          "actor_optim_spec": {
           "lr__choice": [0.0001, 0.0005, 0.001, 0.005, 0.01]
         },
        }
      }]
    }
  },
}