bugfixes and cython

bolero/controller/controller.py

@@ -237,8 +237,8 @@ def episode(self, meta_parameter_keys=(), meta_parameters=()):
 
         return feedbacks
 
-    def episode_with(self, behavior, meta_parameter_keys=[],
-                     meta_parameters=[], record=True):
+    def episode_with(self, behavior, meta_parameter_keys=(),
+                     meta_parameters=(), record=True):
         """Execute a behavior in the environment.
 
         Parameters
@@ -319,7 +319,7 @@ class ContextualController(Controller):
     * test_contexts (array-like) - the upper-level policy will be evaluated in
       these contexts
     """
-    def __init__(self, config={}, environment=None, behavior_search=None,
+    def __init__(self, config=None, environment=None, behavior_search=None,
                  **kwargs):
         super(ContextualController, self).__init__(
             config, environment, behavior_search, **kwargs)

bolero/environment/openaigym.py

@@ -143,7 +143,7 @@ def step_action(self):
         self.done = self.done or done
 
         self.step += 1
-        if self.step >= self.env.spec.timestep_limit:
+        if self.step >= self.env.spec.max_episode_steps:
             self.done = True
 
         if self.log_to_stdout or self.log_to_file:

bolero/environment/optimum_trajectory.py

@@ -86,10 +86,15 @@ def __init__(self,
         self.g = g
         self.execution_time = execution_time
         self.dt = dt
+        task_space_dim = len(x0)
         if obstacles is None:
-            self.obstacles = np.empty((0, len(x0)))
+            self.obstacles = np.empty((0, task_space_dim))
         else:
-            self.obstacles = np.vstack(obstacles)
+            if len(obstacles[:]) % task_space_dim != 0:
+                raise ValueError("Obstacles defined by %d values. A multiple of %d (task space dimension) is required."
+                                 % (len(obstacles[:]), task_space_dim))
+            self.obstacles = np.asarray(obstacles).reshape([-1, task_space_dim])
+            assert self.obstacles.shape[1] == 2
         self.obstacle_dist = obstacle_dist
         self.penalty_start_dist = penalty_start_dist
         self.penalty_goal_dist = penalty_goal_dist

bolero/representation/promp_behavior.py

@@ -290,7 +290,7 @@ def get_n_params(self):
         if self.learn_covariance:
             correlation_coefficients = (
                 len(self.data.covariance_) - random_variables) / 2
-            return 2 * random_variables + correlation_coefficients
+            return int(2 * random_variables + correlation_coefficients)
         else:
             return random_variables
 

bootstrap_bolero.sh

@@ -6,6 +6,13 @@ then
     echo -e "\e[31mPython '$PYTHON' not found.\e[0m"
     exit 1
 fi
+CYTHON_AVAILABLE=1
+`$PYTHON -c "import Cython" 2> /dev/null` || CYTHON_AVAILABLE=0
+if [ CYTHON_AVAILABLE == 0 ];
+then
+        echo "Cython for $PYTHON not available, trying to install it with '$PYTHON -m pip install Cython'"
+        $PYTHON -m pip install Cython
+fi
 echo -e "\e[31mUsing Python: $PYTHON (located at `which $PYTHON`)\e[0m"
 
 # checking minimal dependencies on Ubuntu systems...

examples/optimizer/plot_cmaes.py

@@ -71,7 +71,7 @@ def plot_ellipse(cov, mean, color):
     plt.figure(figsize=(n_generations * 3 / n_rows, 3 * n_rows))
     path = []
     for it in range(n_generations):
-        plt.subplot(n_rows, n_generations / n_rows, it + 1)
+        plt.subplot(n_rows, int(n_generations / n_rows, it + 1))
         plot_objective()
         last_mean = cmaes.mean.copy()
         path.append(last_mean)

examples/optimizer/plot_contextual_optimization.py

@@ -72,7 +72,7 @@ def plot_objective():
 colors = {"C-CMA-ES": "r", "C-REPS": "g"}
 plt.figure(figsize=(n_generations * 3 / n_rows, 3 * n_rows))
 for it in range(n_generations):
-    plt.subplot(n_rows, n_generations / n_rows, it + 1)
+    plt.subplot(n_rows, int(n_generations / n_rows, it + 1))
     plot_objective()
 
     contexts = random_state.rand(n_samples_per_update, 1) * 10.0 - 5.0