refactor(mamba): adjust Mamba2Encoder layer configuration

layers/MambaSeries.py

@@ -23,7 +23,7 @@ class Mamba2Encoder(nn.Module):
         expand=2,
         headdim=64,
         # 堆叠层数
-        n_layers=1,
+        n_layers=2,
     ):
         super().__init__()
         self.patch_num = patch_num
@@ -57,13 +57,3 @@ class Mamba2Encoder(nn.Module):
         u = x.reshape(bs * n_vars, patch_num, self.d_model)  # u: [bs*nvars, patch_num, d_model]
 
         # 3) 通过 n_layers 层 Mamba2 进行建模（在 patch_num 维度上）
-        for m in self.mambas:
-            u = m(u)  # 形状保持 [bs*nvars, patch_num, d_model]
-
-        # 4) 仅取最后一个时间步
-        y_last = u[:, -1, :]  # y_last: [bs*nvars, d_model]
-
-        # 5) 还原回 (bs, nvars, d_model)
-        y_last = y_last.view(bs, n_vars, self.d_model)  # y_last: [bs, nvars, d_model]
-
-        return y_last  # [bs, nvars, d_model]

layers/SeasonPatch.py

@@ -58,7 +58,7 @@ class SeasonPatch(nn.Module):
             self.encoder = Mamba2Encoder(
                 c_in=c_in, patch_num=patch_num, patch_len=patch_len,
                 d_model=d_model, d_state=d_state, d_conv=d_conv,
-                expand=expand, headdim=headdim
+                expand=expand, headdim=headdim, n_layers=n_layers
             )
             # Prediction head（Mamba2 路径用到，输入维度为 d_model）
             self.head = nn.Sequential(