Fix a few more things

supersuit/generic_wrappers/frame_skip.py

@@ -10,12 +10,11 @@
 class frame_skip_gym(gymnasium.Wrapper):
     def __init__(self, env, num_frames):
         super().__init__(env)
-        self.np_random = env.unwrapped.np_random
         self.num_frames = check_transform_frameskip(num_frames)
 
     def step(self, action):
         low, high = self.num_frames
-        num_skips = int(self.np_random.integers(low, high + 1))
+        num_skips = int(self.env.unwrapped.np_random.integers(low, high + 1))
         total_reward = 0.0
 
         for x in range(num_skips):

supersuit/lambda_wrappers/reward_lambda.py

@@ -24,7 +24,7 @@ def reset(self, seed=None, options=None):
         super().reset(seed=seed, options=options)
         self.rewards = {
             agent: self._change_reward_fn(reward)
-            for agent, reward in self.rewards.items()
+            for agent, reward in self.env.rewards.items()  # you don't want to unwrap here, because another reward wrapper might have been applied
         }
         self.__cumulative_rewards = make_defaultdict({a: 0 for a in self.agents})
         self._accumulate_rewards()
@@ -34,7 +34,7 @@ def step(self, action):
         super().step(action)
         self.rewards = {
             agent: self._change_reward_fn(reward)
-            for agent, reward in self.rewards.items()
+            for agent, reward in self.env.rewards.items()  # you don't want to unwrap here, because another reward wrapper might have been applied
         }
         self.__cumulative_rewards[agent] = 0
         self._cumulative_rewards = self.__cumulative_rewards