Spaces:
Sleeping
Sleeping
| from pydantic import BaseModel | |
| class AgentConfig(BaseModel): | |
| state_dim: int = 6 | |
| action_dim: int = 3 | |
| max_action: float = 1.0 | |
| # SAC (Soft Actor-Critic) Hyperparameters | |
| lr_actor: float = 3e-4 | |
| lr_critic: float = 3e-4 | |
| lr_alpha: float = 3e-4 # for temperature tuning | |
| gamma: float = 0.99 | |
| tau: float = 0.005 # Target network update rate | |
| alpha: float = 0.2 # Initial entropy coefficient | |
| buffer_size: int = 100000 | |
| batch_size: int = 256 | |
| # Exploration / Target properties | |
| exploration_steps: int = 5000 # initial random steps | |
| policy_freq: int = 2 | |
| target_entropy: float = -3.0 # -action_dim is a good heuristic | |
| # Weights for BESS multi-objective (used if you want to bias the agent) | |
| # rewards = alpha_ea*r_ea + alpha_fr*r_fr + alpha_ps*r_ps | |