Restructured examples and fixed bugs

fancy_gym/examples/example_replanning.py

@@ -0,0 +1 @@
+# TODO

fancy_gym/examples/example_replanning_envs.py

@@ -6,21 +6,21 @@ def example_run_replanning_env(env_name="fancy_ProDMP/BoxPushingDenseReplan-v0",
     env = gym.make(env_name)
     env.reset(seed=seed)
     for i in range(iterations):
-        done = False
-        while done is False:
+        while True:
             ac = env.action_space.sample()
             obs, reward, terminated, truncated, info = env.step(ac)
             if render:
                 env.render(mode="human")
             if terminated or truncated:
                 env.reset()
+                break
     env.close()
     del env
 
 
 def example_custom_replanning_envs(seed=0, iteration=100, render=True):
     # id for a step-based environment
-    base_env_id = "BoxPushingDense-v0"
+    base_env_id = "fancy/BoxPushingDense-v0"
 
     wrappers = [fancy_gym.envs.mujoco.box_pushing.mp_wrapper.MPWrapper]
 
@@ -38,7 +38,8 @@ def example_custom_replanning_envs(seed=0, iteration=100, render=True):
                         'replanning_schedule': lambda pos, vel, obs, action, t: t % 25 == 0,
                         'condition_on_desired': True}
 
-    env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs=black_box_kwargs,
+    base_env = gym.make(base_env_id)
+    env = fancy_gym.make_bb(env=base_env, wrappers=wrappers, black_box_kwargs=black_box_kwargs,
                             traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
                             phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
                             seed=seed)
@@ -56,10 +57,12 @@ def example_custom_replanning_envs(seed=0, iteration=100, render=True):
     env.close()
     del env
 
-
-if __name__ == "__main__":
+def main(render=False):
     # run a registered replanning environment
-    example_run_replanning_env(env_name="fancy_ProDMP/BoxPushingDenseReplan-v0", seed=1, iterations=1, render=False)
+    example_run_replanning_env(env_name="fancy_ProDMP/BoxPushingDenseReplan-v0", seed=1, iterations=1, render=render)
 
     # run a custom replanning environment
-    example_custom_replanning_envs(seed=0, iteration=8, render=True)
+    example_custom_replanning_envs(seed=0, iteration=8, render=render)
+
+if __name__ == "__main__":
+    main()

fancy_gym/examples/examples_dmc.py

@@ -84,7 +84,8 @@ def example_custom_dmc_and_mp(seed=1, iterations=1, render=True):
     # basis_generator_kwargs = {'basis_generator_type': 'rbf',
     #                           'num_basis': 5
     #                           }
-    env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
+    base_env = gym.make(base_env_id)
+    env = fancy_gym.make_bb(env=base_env, wrappers=wrappers, black_box_kwargs={},
                             traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
                             phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
                             seed=seed)
@@ -114,21 +115,13 @@ def example_custom_dmc_and_mp(seed=1, iterations=1, render=True):
     env.close()
     del env
 
-
-if __name__ == '__main__':
-    # Disclaimer: DMC environments require the seed to be specified in the beginning.
-    # Adjusting it afterwards with env.seed() is not recommended as it does not affect the underlying physics.
-
-    # For rendering DMC
-    # export MUJOCO_GL="osmesa"
-    render = True
-
+def main(render = True):
     # # Standard DMC Suite tasks
     example_dmc("dm_control/fish-swim", seed=10, iterations=1000, render=render)
     #
     # # Manipulation tasks
     # # Disclaimer: The vision versions are currently not integrated and yield an error
-    example_dmc("dm_control/manipulation-reach_site_features", seed=10, iterations=250, render=render)
+    example_dmc("dm_control/reach_site_features", seed=10, iterations=250, render=render)
     #
     # # Gym + DMC hybrid task provided in the MP framework
     example_dmc("dm_control_ProMP/ball_in_cup-catch-v0", seed=10, iterations=1, render=render)
@@ -136,3 +129,20 @@ if __name__ == '__main__':
     # Custom DMC task # Different seed, because the episode is longer for this example and the name+seed combo is
     # already registered above
     example_custom_dmc_and_mp(seed=11, iterations=1, render=render)
+
+    # # Standard DMC Suite tasks
+    example_dmc("dm_control/fish-swim", seed=10, iterations=1000, render=render)
+    #
+    # # Manipulation tasks
+    # # Disclaimer: The vision versions are currently not integrated and yield an error
+    example_dmc("dm_control/reach_site_features", seed=10, iterations=250, render=render)
+    #
+    # # Gym + DMC hybrid task provided in the MP framework
+    example_dmc("dm_control_ProMP/ball_in_cup-catch-v0", seed=10, iterations=1, render=render)
+
+    # Custom DMC task # Different seed, because the episode is longer for this example and the name+seed combo is
+    # already registered above
+    example_custom_dmc_and_mp(seed=11, iterations=1, render=render)
+
+if __name__ == '__main__':
+    main()

fancy_gym/examples/examples_general.py

@@ -85,10 +85,7 @@ def example_async(env_id="fancy/HoleReacher-v0", n_cpu=4, seed=int('533D', 16),
     # do not return values above threshold
     return *map(lambda v: np.stack(v)[:n_samples], buffer.values()),
 
-
-if __name__ == '__main__':
-    render = True
-
+def main(render = True):
     # Basic gym task
     example_general("Pendulum-v1", seed=10, iterations=200, render=render)
 
@@ -100,3 +97,6 @@ if __name__ == '__main__':
 
     # Vectorized multiprocessing environments
     # example_async(env_id="HoleReacher-v0", n_cpu=2, seed=int('533D', 16), n_samples=2 * 200)
+
+if __name__ == '__main__':
+    main()

fancy_gym/examples/examples_metaworld.py

@@ -35,7 +35,7 @@ def example_meta(env_id="fish-swim", seed=1, iterations=1000, render=True):
         if terminated or truncated:
             print(env_id, rewards)
             rewards = 0
-            obs = env.reset()
+            obs = env.reset(seed=seed+i+1)
 
     env.close()
     del env
@@ -81,7 +81,8 @@ def example_custom_meta_and_mp(seed=1, iterations=1, render=True):
     basis_generator_kwargs = {'basis_generator_type': 'rbf',
                               'num_basis': 5
                               }
-    env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
+    base_env = gym.make(base_env_id)
+    env = fancy_gym.make_bb(env=base_env, wrappers=wrappers, black_box_kwargs={},
                             traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
                             phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
                             seed=seed)
@@ -92,14 +93,10 @@ def example_custom_meta_and_mp(seed=1, iterations=1, render=True):
     # It is also possible to change them mode multiple times when
     # e.g. only every nth trajectory should be displayed.
     if render:
-        raise ValueError("Metaworld render interface bug does not allow to render() fixes its interface. "
-                         "A temporary workaround is to alter their code in MujocoEnv render() from "
-                         "`if not offscreen` to `if not offscreen or offscreen == 'human'`.")
-        # TODO: Remove this, when Metaworld fixes its interface.
-        # env.render(mode="human")
+        env.render(mode="human")
 
     rewards = 0
-    obs = env.reset()
+    obs = env.reset(seed=seed)
 
     # number of samples/full trajectories (multiple environment steps)
     for i in range(iterations):
@@ -110,25 +107,23 @@ def example_custom_meta_and_mp(seed=1, iterations=1, render=True):
         if terminated or truncated:
             print(base_env_id, rewards)
             rewards = 0
-            obs = env.reset()
+            obs = env.reset(seed=seed+i+1)
 
     env.close()
     del env
 
-
-if __name__ == '__main__':
-    # Disclaimer: MetaWorld environments require the seed to be specified in the beginning.
-    # Adjusting it afterwards with env.seed() is not recommended as it may not affect the underlying behavior.
-
+def main(render = False):
     # For rendering it might be necessary to specify your OpenGL installation
     # export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so
-    render = False
 
     # # Standard Meta world tasks
     example_meta("metaworld/button-press-v2", seed=10, iterations=500, render=render)
 
     # # MP + MetaWorld hybrid task provided in the our framework
-    example_meta("metaworld_ProMP/ButtonPress-v2", seed=10, iterations=1, render=render)
+    example_meta("metaworld_ProMP/button-press-v2", seed=10, iterations=1, render=render)
     #
     # # Custom MetaWorld task
     example_custom_meta_and_mp(seed=10, iterations=1, render=render)
+
+if __name__ == '__main__':
+    main()

fancy_gym/examples/examples_movement_primitives.py

@@ -28,13 +28,7 @@ def example_mp(env_name="fancy_ProMP/HoleReacher-v0", seed=1, iterations=1, rend
         if render and i % 1 == 0:
             # This renders the full MP trajectory
             # It is only required to call render() once in the beginning, which renders every consecutive trajectory.
-            # Resetting to no rendering, can be achieved by render(mode=None).
-            # It is also possible to change the mode multiple times when
-            # e.g. only every second trajectory should be displayed, such as here
-            # Just make sure the correct mode is set before executing the step.
             env.render(mode="human")
-        else:
-            env.render()
 
         # Now the action space is not the raw action but the parametrization of the trajectory generator,
         # such as a ProMP
@@ -129,7 +123,8 @@ def example_fully_custom_mp(seed=1, iterations=1, render=True):
     # basis_generator_kwargs = {'basis_generator_type': 'rbf',
     #                           'num_basis': 5
     #                           }
-    env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
+    base_env = gym.make(base_env_id)
+    env = fancy_gym.make_bb(env=base_env, wrappers=wrappers, black_box_kwargs={},
                             traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
                             phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
                             seed=seed)
@@ -151,9 +146,7 @@ def example_fully_custom_mp(seed=1, iterations=1, render=True):
             rewards = 0
             obs = env.reset()
 
-
-if __name__ == '__main__':
-    render = False
+def main(render = False):
     # DMP
     example_mp("fancy_DMP/HoleReacher-v0", seed=10, iterations=5, render=render)
 
@@ -172,3 +165,6 @@ if __name__ == '__main__':
 
     # Custom MP
     example_fully_custom_mp(seed=10, iterations=1, render=render)
+
+if __name__ == '__main__':
+    main()

fancy_gym/examples/examples_open_ai.py

@@ -31,6 +31,8 @@ def example_mp(env_name, seed=1, render=True):
             print(returns)
             obs = env.reset()
 
+def main(render=True):
+    example_mp("gym_ProMP/Reacher-v2", render=render)
 
 if __name__ == '__main__':
-    example_mp("gym_ProMP/Reacher-v2")
+    main()