CANDLE Integration

CANDLE functionality is built into EXARL. - Add/modify the learner parameters in ExaRL/config/learner_cfg.json

E.g.:-

{
    "agent": "DQN-v0",
    "env": "ExaLearnCartpole-v1",
    "workflow": "async",
    "n_episodes": 1,
    "n_steps": 10,
    "output_dir": "./exa_results_dir"
}

• Add/modify the agent parameters in ExaRL/exarl/agents/agent_vault/agent_cfg/<AgentName>_<model_type>.json

E.g.:-

{
    "gamma": 0.75,
    "epsilon": 1.0,
    "epsilon_min" : 0.01,
    "epsilon_decay" : 0.999,
    "learning_rate" : 0.001,
    "batch_size" : 32,
    "tau" : 0.5,
    "model_type" : "MLP",
    "dense" : [64, 128],
    "activation" : "relu",
    "optimizer" : "adam",
    "loss" : "mse"
}

Currently, DQN agent takes either MLP or LSTM as model_type. - Add/modify the environment parameters in ExaRL/exarl/envs/env_vault/env_cfg/<EnvName>.json

E.g.:-

{
        "worker_app": "./exarl/envs/env_vault/cpi.py"
}

• Add/modify the workflow parameters in ExaRL/exarl/config/workflow_cfg/<WorkflowName>.json

E.g.:-

{
        "process_per_env": "1"
}

• Please note the agent, environment, and workflow configuration file (json file) name must match the agent, environment, and workflow ID specified in ExaRL/config/learner_cfg.json.

E.g.:- ExaRL/exarl/agents/agent_vault/agent_cfg/DQN-v0_LSTM.json, ExaRL/exarl/envs/env_vault/env_cfg/ExaCartPole-v1.json, and ExaRL/exarl/workflows/workflow_vault/workflow_cfg/async.json