config.json

{
    "total_timesteps": 20000000,
    "policy": "MlpPolicy",
    "policy_kwargs": {
        "log_std_init": -3,
        "activation_fn": "nn.ReLU",
        "net_arch": [256, 256]
    },
    "learning_rate": "linear_schedule(1e-3,5e-4)",
    "batch_size": 256,
    "gamma": 0.99,
    "learning_starts": 10000,
    "buffer_size": 1000000,
    "tau": 0.005,
    "ent_coef": "auto",
    "train_freq": 1,
    "gradient_steps": 1,
    "use_sde": true
}