config.json

{
    "train":{
        "num_epochs": 300,
        "batch_size": 1024,
        "lr": 0.0001,
        "lambda1": 1.0,
        "lambda2": 1.0
    },
    "experiment": {
        "name": "mile",
        "mode": "iterative",
        "validate": {
            "enabled": true,
            "every_n_epochs": 100
        },
        "logging": {
            "terminal_output_to_txt": false,
            "log_tb": false,
            "log_wandb": true
        },
        "save": {
            "enabled": true,
            "every_n_epochs": 0,
            "on_best_validation": false,
            "on_best_rollout_return": false,
            "on_best_rollout_success_rate": true,
            "outdir": "output_dir"
        },
        "rollout": {
            "enabled": true,
            "n_episodes": 30,
            "every_n_epochs": 100
        },
        "env_name": "peg-insert-side-v2",
        "dataset_path": "path_to_offline_intervention_dataset",
        "policy_type": "sac",
        "policy_path": "./trained_models/initial_policy",
        "mental_model_type": "bc",
        "gt_mental_model_path": "./trained_models/gt_mental_model",
        "intervention_policy_path": "./trained_models/expert_policy",
        "use_warm_start": true,
        "warm_start_path": "./trained_models/warm_started_mental_model",
        "num_rounds": 20,
        "episodes_per_round": 1,
        "include_offline_dataset": false,
        "offline_dataset_path": "path_to_initial_offline_demonstration_dataset"
    }
}