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

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

vllm/model_executor/models/qwen_vl.py

@@ -25,6 +25,7 @@ from transformers.tokenization_utils_base import TextInput
 from vllm.config import VllmConfig
 from vllm.config.multimodal import BaseDummyOptions
 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,
     ReplicatedLinear,
@@ -333,7 +334,7 @@ class VisionTransformer(nn.Module):
         patch_height, patch_width = self.patch_size = (patch_size, patch_size)
         self.grid_size = (image_height // patch_height, image_width // patch_width)
         self.output_dim = output_dim
-        self.conv1 = nn.Conv2d(
+        self.conv1 = Conv2dLayer(
             in_channels=3,
             out_channels=width,
             kernel_size=patch_size,