diff --git a/tianshou/policy/modelfree/sac.py b/tianshou/policy/modelfree/sac.py
index d701aca..3a55fba 100644
--- a/tianshou/policy/modelfree/sac.py
+++ b/tianshou/policy/modelfree/sac.py
@@ -87,6 +87,7 @@ class SACPolicy(DDPGPolicy):
         act = y * self._action_scale + self._action_bias
         log_prob = dist.log_prob(x) - torch.log(
             self._action_scale * (1 - y.pow(2)) + self.__eps)
+        log_prob = torch.unsqueeze(torch.sum(log_prob, 1), 1)
         act = act.clamp(self._range[0], self._range[1])
         return Batch(
             logits=logits, act=act, state=h, dist=dist, log_prob=log_prob)