428 lines
14 KiB
Python
428 lines
14 KiB
Python
"""Simulate a trained policy in the MuJoCo viewer.
|
|
|
|
Uses Hydra to load the same BrittleStarConfig that was used during training.
|
|
Override settings via CLI, e.g.:
|
|
python scripts/simulate.py morphology=3_arms
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import itertools
|
|
import time
|
|
from pathlib import Path
|
|
from typing import Any
|
|
|
|
import flax
|
|
import hydra
|
|
import jax
|
|
import jax.numpy as jnp
|
|
import numpy as np
|
|
from omegaconf import DictConfig, OmegaConf
|
|
|
|
from brittle_star_project import BrittleStarEnv, BrittleStarEnvFactory
|
|
from brittle_star_project.configs.main_config import BrittleStarConfig
|
|
from brittle_star_project.configs.register_configs import register_configs
|
|
|
|
_ALLOWED_OBS_KEYS = {
|
|
"joint_position",
|
|
"joint_velocity",
|
|
"joint_actuator_force",
|
|
"actuator_force",
|
|
"disk_position",
|
|
"disk_rotation",
|
|
"disk_linear_velocity",
|
|
"disk_angular_velocity",
|
|
"unit_xy_direction_to_target",
|
|
"xy_distance_to_target",
|
|
}
|
|
|
|
def _transform_obs_dict(obs_dict: dict[str, Any]) -> jnp.ndarray:
|
|
"""Flatten the env's observation dict into a 1D vector.
|
|
|
|
Matches training behavior:
|
|
- only includes keys in _ALLOWED_OBS_KEYS
|
|
- iterates keys in sorted order for stable layout
|
|
- skips empty arrays
|
|
"""
|
|
parts: list[jnp.ndarray] = []
|
|
for key in sorted(obs_dict.keys()):
|
|
if key not in _ALLOWED_OBS_KEYS:
|
|
continue
|
|
arr = jnp.asarray(obs_dict[key])
|
|
if arr.size == 0:
|
|
continue
|
|
parts.append(arr.reshape((-1,)))
|
|
|
|
if not parts:
|
|
return jnp.zeros((0,), dtype=jnp.float32)
|
|
return jnp.concatenate(parts, axis=0)
|
|
|
|
|
|
# A minimal policy class to load a CleanRL/Flax checkpoint and run inference.
|
|
class CleanRLPPOPolicy:
|
|
def __init__(
|
|
self,
|
|
*,
|
|
sensor_params: Any,
|
|
actor_params: Any,
|
|
action_dim: int,
|
|
) -> None:
|
|
from brittle_star_project.MLPs.mlps import Actor, GenericDenseLayersWithActivation
|
|
|
|
hidden_dim = int(sensor_params["params"]["Dense_0"]["kernel"].shape[1])
|
|
|
|
self._sensor = GenericDenseLayersWithActivation(layer_sizes=[hidden_dim, hidden_dim])
|
|
self._actor = Actor(action_dim=action_dim)
|
|
self._sensor_apply = jax.jit(self._sensor.apply)
|
|
self._actor_apply = jax.jit(self._actor.apply)
|
|
self._params = {
|
|
"sensor_params": sensor_params,
|
|
"actor_params": actor_params,
|
|
}
|
|
|
|
@staticmethod
|
|
def load(
|
|
path: Path,
|
|
*,
|
|
action_dim: int,
|
|
) -> "CleanRLPPOPolicy":
|
|
def _get_index(container: Any, idx: int) -> Any:
|
|
if isinstance(container, (list, tuple)):
|
|
return container[idx]
|
|
if isinstance(container, dict):
|
|
return container.get(idx, container.get(str(idx)))
|
|
raise KeyError(idx)
|
|
|
|
def _looks_like_indexed_dict(container: Any) -> bool:
|
|
return (
|
|
isinstance(container, dict)
|
|
and container
|
|
and all(str(k).isdigit() for k in container.keys())
|
|
)
|
|
|
|
def _parse_checkpoint(restored_obj: Any) -> tuple[Any, Any, Any, Any, Any]:
|
|
"""Extract checkpoint parts.
|
|
|
|
Returns (config_dict, sensor_params, actor_params, critic_params,
|
|
feature_extractor_params).
|
|
|
|
PPOTrainer saves:
|
|
flax.serialization.to_bytes([
|
|
config_dict,
|
|
[sensor_params, actor_params, critic_params, feature_extractor_params],
|
|
])
|
|
|
|
msgpack_restore() may restore lists as dicts keyed by string indices
|
|
("0", "1", ...), so we accept both shapes.
|
|
"""
|
|
|
|
cfg_part: Any | None = None
|
|
params_part: Any = restored_obj
|
|
|
|
if isinstance(restored_obj, (list, tuple)) and len(restored_obj) >= 2:
|
|
cfg_part = restored_obj[0]
|
|
params_part = restored_obj[1]
|
|
elif _looks_like_indexed_dict(restored_obj) and (
|
|
"0" in restored_obj or "1" in restored_obj
|
|
):
|
|
cfg_part = restored_obj.get("0", restored_obj.get(0))
|
|
params_part = restored_obj.get("1", restored_obj.get(1))
|
|
|
|
if _looks_like_indexed_dict(params_part):
|
|
sensor_params = _get_index(params_part, 0)
|
|
actor_params = _get_index(params_part, 1)
|
|
critic_params = _get_index(params_part, 2)
|
|
feature_extractor_params = _get_index(params_part, 3)
|
|
if sensor_params is None or actor_params is None:
|
|
raise ValueError("Missing required params in checkpoint")
|
|
return (
|
|
cfg_part,
|
|
sensor_params,
|
|
actor_params,
|
|
critic_params,
|
|
feature_extractor_params,
|
|
)
|
|
|
|
if isinstance(params_part, (list, tuple)) and len(params_part) >= 2:
|
|
sensor_params = params_part[0]
|
|
actor_params = params_part[1]
|
|
critic_params = params_part[2] if len(params_part) >= 3 else None
|
|
feature_extractor_params = params_part[3] if len(params_part) >= 4 else None
|
|
return (
|
|
cfg_part,
|
|
sensor_params,
|
|
actor_params,
|
|
critic_params,
|
|
feature_extractor_params,
|
|
)
|
|
|
|
raise ValueError(
|
|
f"Unexpected checkpoint structure in {path}. "
|
|
"Expected [config_dict, [sensor_params, actor_params, critic_params, "
|
|
"feature_extractor_params]] or an equivalent dict-indexed variant."
|
|
)
|
|
|
|
payload = path.read_bytes()
|
|
restored = flax.serialization.msgpack_restore(payload)
|
|
_cfg_dict, sensor_params, actor_params, _critic_params, _feature_extractor_params = (
|
|
_parse_checkpoint(restored)
|
|
)
|
|
|
|
return CleanRLPPOPolicy(
|
|
sensor_params=sensor_params,
|
|
actor_params=actor_params,
|
|
action_dim=action_dim,
|
|
)
|
|
|
|
def act(self, *, observations: dict[str, Any]) -> np.ndarray:
|
|
obs = _transform_obs_dict(observations)
|
|
hidden = self._sensor_apply(self._params["sensor_params"], obs)
|
|
mean, _log_std = self._actor_apply(self._params["actor_params"], hidden)
|
|
|
|
# Always evaluate with the actor mean.
|
|
# (Sampling adds exploration noise, which is useful for training but not for evaluation.)
|
|
return np.asarray(mean, dtype=np.float32).ravel()
|
|
|
|
|
|
def _get_observations(state: Any) -> dict[str, Any] | None:
|
|
return getattr(state, "observations", None)
|
|
|
|
|
|
def _get_xy_distance_to_target(observations: dict[str, Any]) -> float | None:
|
|
return float(np.asarray(observations["xy_distance_to_target"]).reshape(-1)[0])
|
|
|
|
|
|
def _target_reached(*, state: Any) -> bool:
|
|
return bool(getattr(state, "terminated", False))
|
|
|
|
|
|
def _rollout_one_episode_headless(
|
|
*,
|
|
env: BrittleStarEnv,
|
|
policy: CleanRLPPOPolicy,
|
|
seed: int,
|
|
max_steps: int,
|
|
) -> tuple[float, int, bool, float | None]:
|
|
"""Run one rollout up to max_steps.
|
|
|
|
Returns (return, length, reached_target, final_xy_dist).
|
|
|
|
Note: In the MJC backend, the raw env reward can be 0.0; we compute a simple
|
|
progress reward based on xy_distance_to_target.
|
|
"""
|
|
|
|
state = env.reset(seed=seed)
|
|
|
|
ep_return = 0.0
|
|
observations = _get_observations(state)
|
|
prev_dist = _get_xy_distance_to_target(observations)
|
|
reached_target = _target_reached(state=state)
|
|
|
|
steps = 0
|
|
for _ in range(int(max_steps)):
|
|
action = policy.act(observations=observations)
|
|
|
|
nu = int(state.mj_model.nu)
|
|
if nu > 0 and action.shape != (nu,):
|
|
raise ValueError(f"Policy returned action shape {action.shape}, expected ({nu},)")
|
|
|
|
state = env.step(state=state, action=action)
|
|
steps += 1
|
|
|
|
observations = _get_observations(state)
|
|
cur_dist = _get_xy_distance_to_target(observations)
|
|
if prev_dist is not None and cur_dist is not None:
|
|
ep_return += prev_dist - cur_dist
|
|
prev_dist = cur_dist
|
|
|
|
reached_target = _target_reached(state=state)
|
|
if reached_target:
|
|
break
|
|
|
|
final_dist = _get_xy_distance_to_target(observations)
|
|
return ep_return, steps, reached_target, final_dist
|
|
|
|
|
|
def _run_one_episode_viewer(
|
|
*,
|
|
env: BrittleStarEnv,
|
|
policy: CleanRLPPOPolicy,
|
|
seed: int,
|
|
state: Any,
|
|
control_dt: float,
|
|
max_steps: int | None,
|
|
) -> None:
|
|
import mujoco.viewer
|
|
|
|
model = state.mj_model
|
|
data = state.mj_data
|
|
|
|
_ = int(seed)
|
|
episode_return = 0.0
|
|
observations = _get_observations(state)
|
|
prev_dist = _get_xy_distance_to_target(observations)
|
|
reached_target = _target_reached(state=state)
|
|
|
|
steps = 0
|
|
# Use the viewer as a context manager to avoid GLX teardown races.
|
|
with mujoco.viewer.launch_passive(model, data) as viewer:
|
|
step_iter = range(int(max_steps)) if max_steps is not None else itertools.count()
|
|
for _step_idx in step_iter:
|
|
if not viewer.is_running():
|
|
break
|
|
step_start = time.time()
|
|
|
|
action = policy.act(observations=observations or {})
|
|
if model.nu > 0 and action.shape != (int(model.nu),):
|
|
raise ValueError(
|
|
f"Policy returned action shape {action.shape}, expected ({int(model.nu)},)"
|
|
)
|
|
|
|
# The passive viewer runs a GUI thread; protect MuJoCo state mutation.
|
|
with viewer.lock():
|
|
state = env.step(state=state, action=action)
|
|
|
|
if not viewer.is_running():
|
|
break
|
|
viewer.sync()
|
|
|
|
steps += 1
|
|
|
|
observations = _get_observations(state)
|
|
cur_dist = _get_xy_distance_to_target(observations)
|
|
if prev_dist is not None and cur_dist is not None:
|
|
episode_return += prev_dist - cur_dist
|
|
prev_dist = cur_dist
|
|
|
|
reached_target = _target_reached(state=state)
|
|
if reached_target:
|
|
break
|
|
|
|
remaining = control_dt - (time.time() - step_start)
|
|
if remaining > 0:
|
|
time.sleep(remaining)
|
|
|
|
dist = _get_xy_distance_to_target(observations)
|
|
dist_str = "n/a" if dist is None else f"{dist:.3f}"
|
|
print(
|
|
"episode done: "
|
|
f"return={episode_return:.6f}, len={steps}, "
|
|
f"target_reached={reached_target}, final_xy_dist={dist_str}"
|
|
)
|
|
|
|
|
|
def _infer_checkpoint_obs_dim(policy: CleanRLPPOPolicy) -> int | None:
|
|
"""Best-effort read of the first Dense kernel input dim (obs dim)."""
|
|
|
|
try:
|
|
kernel = policy._params["sensor_params"]["params"]["Dense_0"]["kernel"]
|
|
return int(getattr(kernel, "shape")[0])
|
|
except Exception:
|
|
return None
|
|
|
|
|
|
@hydra.main(config_path="../configs", config_name="main_config", version_base="1.3")
|
|
def main(dict_cfg: DictConfig) -> None:
|
|
# Convert DictConfig to structured dataclass, ensuring the root schema is applied.
|
|
config: BrittleStarConfig = OmegaConf.to_object(
|
|
OmegaConf.merge(OmegaConf.structured(BrittleStarConfig), dict_cfg)
|
|
)
|
|
|
|
backend = config.simulation.backend
|
|
seed = int(config.experiment.seed)
|
|
|
|
model_path_str = config.simulation.model_path
|
|
if model_path_str is None:
|
|
raise ValueError(
|
|
"simulation.model_path must be set to a .flax checkpoint (e.g. final_model.flax)"
|
|
)
|
|
|
|
# Hydra chdir changes CWD; resolve relative paths relative to the invocation.
|
|
model_path = Path(hydra.utils.to_absolute_path(model_path_str))
|
|
if model_path.suffix != ".flax":
|
|
raise ValueError(f"Expected a '.flax' checkpoint, got '{model_path.name}'.")
|
|
|
|
# ======= ENVIRONMENT SETUP =======
|
|
factory = BrittleStarEnvFactory()
|
|
raw_env = factory.create_environment(
|
|
backend,
|
|
config.morphology,
|
|
config.arena,
|
|
config.environment,
|
|
)
|
|
env = BrittleStarEnv(
|
|
raw_env,
|
|
backend=backend,
|
|
config=config.environment,
|
|
morphology_config=config.morphology,
|
|
)
|
|
|
|
state0 = env.reset(seed=seed)
|
|
|
|
# ======= MODEL SETUP =======
|
|
nu = int(state0.mj_model.nu)
|
|
policy = CleanRLPPOPolicy.load(model_path, action_dim=nu)
|
|
|
|
# Helpful early failure when configs don't match the checkpoint.
|
|
observations0 = _get_observations(state0)
|
|
env_obs_dim = int(_transform_obs_dict(observations0 or {}).shape[0])
|
|
ckpt_obs_dim = _infer_checkpoint_obs_dim(policy)
|
|
|
|
if ckpt_obs_dim is not None and ckpt_obs_dim != env_obs_dim:
|
|
raise ValueError(
|
|
"Checkpoint/env mismatch: "
|
|
f"checkpoint expects obs_dim={ckpt_obs_dim}, env provides obs_dim={env_obs_dim}. "
|
|
"Use the same Hydra config (morphology/arena/environment) "
|
|
"that was used during training."
|
|
)
|
|
|
|
# ======= SIMULATION =======
|
|
headless = bool(config.simulation.headless)
|
|
max_steps = config.simulation.max_steps
|
|
|
|
if headless:
|
|
if max_steps is None:
|
|
raise ValueError("simulation.max_steps is required when simulation.headless=true")
|
|
max_steps_i = int(max_steps)
|
|
if max_steps_i <= 0:
|
|
raise ValueError("simulation.max_steps must be > 0")
|
|
|
|
ep_return, ep_len, reached_target, final_dist = _rollout_one_episode_headless(
|
|
env=env,
|
|
policy=policy,
|
|
seed=seed,
|
|
max_steps=max_steps_i,
|
|
)
|
|
final_dist_str = "n/a" if final_dist is None else f"{final_dist:.3f}"
|
|
print(
|
|
"episode done: "
|
|
f"return={ep_return:.6f}, len={ep_len}, "
|
|
f"target_reached={reached_target}, final_xy_dist={final_dist_str}"
|
|
)
|
|
else:
|
|
if max_steps is not None:
|
|
max_steps_i = int(max_steps)
|
|
if max_steps_i <= 0:
|
|
raise ValueError("simulation.max_steps must be > 0")
|
|
max_steps_val: int | None = max_steps_i
|
|
else:
|
|
max_steps_val = None
|
|
|
|
model_dt = float(state0.mj_model.opt.timestep)
|
|
control_dt = model_dt * float(config.environment.num_physics_steps_per_control_step)
|
|
|
|
_run_one_episode_viewer(
|
|
env=env,
|
|
policy=policy,
|
|
seed=seed,
|
|
state=state0,
|
|
control_dt=control_dt,
|
|
max_steps=max_steps_val,
|
|
)
|
|
|
|
env.close()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
register_configs()
|
|
main()
|