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

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

vllm/model_executor/models/idefics2_vision_model.py

@@ -30,6 +30,7 @@ from transformers.models.idefics2.configuration_idefics2 import (
 from vllm.attention.layer import MultiHeadAttention
 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,
@@ -60,7 +61,7 @@ class Idefics2VisionEmbeddings(nn.Module):
         self.embed_dim = config.hidden_size
         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,