将Tensor改为tensor

Tensor与tensor的问题，规范化tensor使用。

Aimbot-PPO-Python/Pytorch/MultiNN-PPO.py

@@ -112,7 +112,7 @@ if __name__ == "__main__":
                 with torch.no_grad():
                     # predict actions
                     action, dis_logprob, _, con_logprob, _, value = agent.get_actions_value(
-                        torch.Tensor(state).to(device)
+                        torch.tensor(state,dtype=torch.float32).to(device)
                     )
                     value = value.flatten()
 

Aimbot-PPO-Python/Pytorch/aimemory.py

@@ -61,7 +61,7 @@ class PPOMem:
             thisRewardBF = (np.asarray(thisRewardBF) + (remainTime * self.result_broadcast_ratio)).tolist()
         else:
             print("!!!!!DIDNT GET RESULT REWARD!!!!!!", rewardBF[-1])
-        return torch.Tensor(thisRewardBF).to(self.device)
+        return torch.tensor(thisRewardBF,dtype=torch.float32).to(self.device)
 
     def save_memories(
         self,
@@ -101,10 +101,10 @@ class PPOMem:
                 thisRewardsTensor = self.broad_cast_end_reward(self.rewards_bf[i], remainTime)
                 adv, rt = agent.gae(
                     rewards=thisRewardsTensor,
-                    dones=torch.Tensor(self.dones_bf[i]).to(self.device),
+                    dones=torch.tensor(self.dones_bf[i],dtype=torch.float32).to(self.device),
                     values=torch.tensor(self.values_bf[i]).to(self.device),
                     next_obs=torch.tensor(next_state[i]).to(self.device).unsqueeze(0),
-                    next_done=torch.Tensor([next_done[i]]).to(self.device),
+                    next_done=torch.tensor([next_done[i]],dtype=torch.float32).to(self.device),
                 )
                 # send memories to training datasets
                 self.obs[roundTargetType] = torch.cat((self.obs[roundTargetType], torch.tensor(np.array(self.ob_bf[i])).to(self.device)), 0)

Aimbot-PPO-Python/Pytorch/arguments.py

@@ -34,7 +34,7 @@ BROADCASTREWARD = False
 ANNEAL_LEARNING_RATE = True
 CLIP_VLOSS = True
 NORM_ADV = False
-TRAIN = False
+TRAIN = True
 SAVE_MODEL = False
 WANDB_TACK = False
 LOAD_DIR = None

Aimbot-PPO-Python/Pytorch/ppoagent.py

@@ -275,8 +275,8 @@ class PPOAgent(nn.Module):
             self,
             rewards: torch.Tensor,
             dones: torch.Tensor,
-            values: torch.tensor,
+            values: torch.Tensor,
-            next_obs: torch.tensor,
+            next_obs: torch.Tensor,
             next_done: torch.Tensor,
     ) -> tuple:
         # GAE