Replace with max-safety weights (start 1 of the 6-run frontier): mean 231.2, landings 99.67%, crashes 0.25% - 23x fewer crashes than the teacher
bb8222a verified | { | |
| "architectures": ["TinyLunarPolicy"], | |
| "model_type": "tiny-lunar-policy", | |
| "auto_map": { | |
| "AutoConfig": "configuration_tiny.TinyLunarConfig", | |
| "AutoModel": "modeling_tiny.TinyLunarPolicy" | |
| }, | |
| "obs_dim": 8, | |
| "n_actions": 4, | |
| "bias": false, | |
| "n_params": 32, | |
| "variant": "max-safety", | |
| "metrics_eval_60x100": { | |
| "mean_return": 231.227, | |
| "std_over_seeds": 3.394, | |
| "seeds_with_mean_ge_200": "60/60", | |
| "landing_rate": 0.9967, | |
| "crash_rate": 0.0025, | |
| "episodes": 6000 | |
| }, | |
| "env_id": "LunarLander-v3", | |
| "observation_names": [ | |
| "x", | |
| "y", | |
| "vx", | |
| "vy", | |
| "angle", | |
| "angular_velocity", | |
| "leg_left_contact", | |
| "leg_right_contact" | |
| ], | |
| "action_meanings": [ | |
| "do nothing", | |
| "fire left orientation engine", | |
| "fire main engine", | |
| "fire right orientation engine" | |
| ], | |
| "torch_dtype": "float32" | |
| } | |