Remove more mujoco code

src/dep/eval.py

@@ -201,73 +201,6 @@ def eval_checkpoint(
     }
 
 
-def eval_from_mujoco_state(
-    mujoco_state_path: str,
-    agent_alg: str = "ppo",
-    render: bool = False,
-    num_steps: int = 10000,
-    wandb_project: str = "pbrl-defense",
-    wandb_group: str = "dev",
-    env_name="mpe",
-    scenario_name="simple_push",
-    artifact_dir: Union[str, Path] = Path("/scratch/pavel/out/tmp"),
-    video_dir=None,
-    local_mode: bool = False,
-) -> dict:
-    artifact_dir = Path(artifact_dir)
-    wandb.init(project=wandb_project, group=wandb_group, dir=artifact_dir)
-
-    ray.init(local_mode=local_mode)
-
-    trainer_class = trainer_cls_from_str(agent_alg)
-
-    mujoco_state_file = Path(mujoco_state_path).open(mode="rb")
-    mujoco_state = pickle.load(mujoco_state_file)
-
-    _ = init_env(
-        env_name=env_name,
-        scenario_name=scenario_name,
-        max_steps=25,
-        mujoco_state=mujoco_state,
-    )
-
-    update_config = _update_config_for_eval(
-        {
-            "env_config": {"scenario_name": scenario_name},
-            "num_workers": 0,
-            "evaluation_num_workers": 0,
-            "custom_eval_function": None,
-            "evaluation_duration": 1,
-        },
-        scenario_name,
-    )
-
-    trainer = trainer_class(
-        env="current-env", config=update_config, logger_creator=noop_logger_creator
-    )
-    # trainer = restore_trainer_from_path(file, scenario_name, trainer_class, config_update=update_config)
-
-    # Now the trainer object should have the agent0 weights in policy0 and agent1 weights in policy1
-    saver = InMemoryRolloutSaver()
-
-    # Suppress command line output because rollout() is a bit too verbose for my taste
-    with contextlib.redirect_stdout(None):
-        rollout(
-            trainer,
-            None,  # Parameter is unused
-            num_steps=num_steps,
-            saver=saver,
-            no_render=not render,
-            video_dir=video_dir,
-        )
-
-    ray.shutdown()
-
-    return {
-        "mean_rewards": saver.mean_rewards,
-    }
-
-
 def _update_config_for_eval(config, scenario_name):
     """Updates the config for evaluation. Code from rllib/rollout.py"""
 
@@ -299,9 +232,5 @@ def _update_config_for_eval(config, scenario_name):
 
 if __name__ == "__main__":
     fire.Fire(
-        {
-            "checkpoint": eval_checkpoint,
-            "mujoco-state": eval_from_mujoco_state,
-            "multi": multi_eval,
-        }
+        {"checkpoint": eval_checkpoint, "multi": multi_eval,}
     )