add more high and low with action_bin

opendilab · Apr 15, 2024 · 81bea50 · 81bea50
1 parent 33554e7
commit 81bea50
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/ding/policy/qtransformer.py b/ding/policy/qtransformer.py
@@ -273,9 +273,9 @@ def _init_learn(self) -> None:
             update_type="momentum",
             update_kwargs={"theta": self._cfg.learn.target_theta},
         )
-        self._low = np.array([-1, -1, -1])
-        self._high = np.array([1, 1, 1])
         self._action_bin = self._cfg.model.action_bins
+        self._low = np.full(self._cfg.model.num_actions, -1)
+        self._high = np.full(self._cfg.model.num_actions, 1)
         self._action_values = np.array(
             [
                 np.linspace(min_val, max_val, self._action_bin)