[Models] Replace all nn.Conv2d with vLLM's Conv2dLayer (#28842)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>

vllm/model_executor/models/midashenglm.py

@@ -39,6 +39,7 @@ from vllm.config import VllmConfig
 from vllm.config.multimodal import BaseDummyOptions
 from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import get_act_fn
+from vllm.model_executor.layers.conv import Conv2dLayer
 from vllm.model_executor.layers.linear import (
     ColumnParallelLinear,
     QKVParallelLinear,
@@ -120,7 +121,7 @@ class AudioPatchEmbed(nn.Module):
         self.num_patches = self.grid_size[0] * self.grid_size[1]
         self.flatten = flatten
 
-        self.proj = nn.Conv2d(
+        self.proj = Conv2dLayer(
             in_chans,
             embed_dim,
             kernel_size=self.patch_size,