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

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

vllm/model_executor/models/paddleocr_vl.py

@@ -45,6 +45,7 @@ from vllm.config.multimodal import BaseDummyOptions
 from vllm.distributed import parallel_state
 from vllm.distributed import utils as dist_utils
 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,
@@ -419,7 +420,7 @@ class SiglipVisionEmbeddings(nn.Module):
         self.image_size = config.image_size
         self.patch_size = config.patch_size
 
-        self.patch_embedding = nn.Conv2d(
+        self.patch_embedding = Conv2dLayer(
             in_channels=config.num_channels,
             out_channels=self.embed_dim,
             kernel_size=self.patch_size,