diff --git a/model/model.py b/model/model.py
index 8e873cb..848cf8a 100644
--- a/model/model.py
+++ b/model/model.py
@@ -221,7 +221,6 @@ class MOEFeedForward(nn.Module):
         x = x.view(-1, x.shape[-1])
         flat_topk_idx = topk_idx.view(-1)
         if self.training:
-            # 训练模式下，重复输入数据
             x = x.repeat_interleave(self.config.num_experts_per_tok, dim=0)
             y = torch.empty_like(x, dtype=torch.float16)
             for i, expert in enumerate(self.experts):
@@ -229,7 +228,6 @@ class MOEFeedForward(nn.Module):
             y = (y.view(*topk_weight.shape, -1) * topk_weight.unsqueeze(-1)).sum(dim=1)
             y = y.view(*orig_shape)
         else:
-            # 推理模式下，只选择最优专家
             y = self.moe_infer(x, flat_topk_idx, topk_weight.view(-1, 1)).view(*orig_shape)
         if self.config.n_shared_experts is not None:
             y = y + self.shared_experts(identity)