updates

alr_envs/classic_control/hole_reacher.py

@@ -35,7 +35,6 @@ class HoleReacher(gym.Env):
         self._angle_velocity = None
         self.start_pos = np.hstack([[np.pi/2], np.zeros(self.num_links - 1)])
         self.start_vel = np.zeros(self.num_links)
-        self.weight_matrix_scale = 50  # for the holereacher, the dmp weights become quite large compared to the values of the goal attractor. this scaling is to ensure they are on similar scale for the optimizer
 
         self.dt = 0.01
         self.time_limit = 2
@@ -68,6 +67,9 @@ class HoleReacher(gym.Env):
     def end_effector(self):
         return self._joints[self.num_links].T
 
+    def configure(self, context):
+        pass
+
     def reset(self):
         self._joint_angles = self.start_pos
         self._angle_velocity = self.start_vel

alr_envs/classic_control/utils.py

@@ -62,7 +62,8 @@ def make_holereacher_env(rank, seed=0):
                              learn_goal=True,
                              alpha_phase=2,
                              start_pos=_env.start_pos,
-                             policy_type="velocity"
+                             policy_type="velocity",
+                             weights_scale=100,
                              )
         _env.seed(seed + rank)
         return _env