polish(pu): polish config

lzero/mcts/buffer/game_buffer.py

@@ -564,15 +564,10 @@ def _push_game_segment(self, data: Any, meta: Optional[dict] = None) -> None:
             # print(f'valid_len is {valid_len}')
 
         if meta['priorities'] is None:
-            # try:
             if self.game_segment_buffer:
                 max_prio = self.game_pos_priorities.max() if len(self.game_pos_priorities) > 0 else 1
             else:
                 max_prio = 1
-            # except Exception as e:
-            #     print(e)
-            #     print(f'self.game_pos_priorities:{self.game_pos_priorities}')
-            #     print(f'self.game_segment_buffer:{self.game_segment_buffer}')
 
             # if no 'priorities' provided, set the valid part of the new-added game history the max_prio
             self.game_pos_priorities = np.concatenate((self.game_pos_priorities, [max_prio for _ in range(valid_len)] + [0. for _ in range(valid_len, data_length)]))

lzero/model/common.py

@@ -562,7 +562,6 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
             # for atari 64,8,8 = 4096 -> 768
             x = self.sim_norm(x)
 
-
         return x
 
 

lzero/model/sampled_muzero_model.py

@@ -3,14 +3,11 @@
 
 import torch
 import torch.nn as nn
-from ding.model.common import ReparameterizationHead
-from ding.torch_utils import MLP
 from ding.utils import MODEL_REGISTRY, SequenceType
 
 from .common import MZNetworkOutput, RepresentationNetwork
-from .sampled_efficientzero_model import PredictionNetwork
-
 from .muzero_model import DynamicsNetwork
+from .sampled_efficientzero_model import PredictionNetwork
 from .utils import renormalize