fix some possible int/float zero issues with previous action and others

CoyoteObs.py

@@ -93,7 +93,7 @@ def __init__(self, tick_skip=8, team_size=3, expanding: bool = True, extra_boost
         if self.doubletap_indicator:
             self.floor_bounce = False
             self.backboard_bounce = False
-            self.prev_ball_vel = np.asarray([0] * 3)
+            self.prev_ball_vel = np.asarray([0.] * 3)
         assert not (selector_infinite_boost is not None and not selector)
         self.n = 0
         self.selector_infinite_boost = selector_infinite_boost

CoyoteParser.py

@@ -668,7 +668,7 @@ def __init__(self, obs_info=None):
         self._lookup_table = self.make_lookup_table(len(self.models))
         # self.prev_action = None
         # self.prev_model = None
-        self.prev_actions = np.asarray([[0] * 8] * 8)
+        self.prev_actions = np.asarray([[0.] * 8] * 8)
         self.prev_model_actions = np.asarray([-1] * 6)
 
     @staticmethod

learner_gp.py

@@ -52,10 +52,10 @@
         ent_coef=0.01,
     )
 
-    run_id = "gp_run_2v2_3.54"
+    run_id = "gp_run_2v2_3.55"
     wandb.login(key=os.environ["WANDB_KEY"])
     logger = wandb.init(dir="./wandb_store",
-                        name="GP_Run_2v2_3.54",
+                        name="GP_Run_2v2_3.55",
                         project="Opti",
                         entity="kaiyotech",
                         id=run_id,
@@ -154,7 +154,7 @@
         disable_gradient_logging=True,
     )
 
-    alg.load("GP_saves/Opti_1683132125.3858902/Opti_42240/checkpoint.pt")
+    alg.load("GP_saves/Opti_1689183465.5270734/Opti_42370/checkpoint.pt")
 
     alg.agent.optimizer.param_groups[0]["lr"] = logger.config.actor_lr
     alg.agent.optimizer.param_groups[1]["lr"] = logger.config.critic_lr