env.step wants 1D array, squeeze action

src/train.py

@@ -102,7 +102,7 @@ def run_episode(env, policy, scaler, animate=False):
         observes.append(obs)
         action = policy.sample(obs).reshape((1, -1)).astype(np.float64)
         actions.append(action)
-        obs, reward, done, _ = env.step(action)
+        obs, reward, done, _ = env.step(np.squeeze(action))
         if not isinstance(reward, float):
             reward = np.asscalar(reward)
         rewards.append(reward)