[ET-VK] Include FuseDequantLinearPass() in vulkan_preprocess

Pull Request resolved: #6168

## Context

Include `FuseDequantLinearPass` as a part of `vulkan_preprocess`, so that fusing the quant/dequant nodes added by `VulkanQuantizer` can be done as part of the lowering process.
ghstack-source-id: 247613964
@exported-using-ghexport

Differential Revision: [D64249613](https://our.internmc.facebook.com/intern/diff/D64249613/)

Author: SS-JIA

backends/vulkan/TARGETS

@@ -25,6 +25,7 @@ runtime.python_library(
         "//executorch/backends/transforms:addmm_mm_to_linear",
         "//executorch/backends/transforms:fuse_batch_norm_with_conv",
         "//executorch/backends/transforms:fuse_conv_with_clamp",
+        "//executorch/backends/transforms:fuse_dequant_linear",
         "//executorch/backends/transforms:fuse_view_copy",
         "//executorch/backends/transforms:mean_to_sum_div",
         "//executorch/backends/transforms:remove_clone_ops",

backends/vulkan/partitioner/supported_ops.py

@@ -45,6 +45,13 @@ def __contains__(self, op):
 
 PRIM_OPS = [
     operator.getitem,
+    # Quantization related ops will be fused via graph passes
+    exir_ops.edge.quantized_decomposed.quantize_per_channel.default,
+    exir_ops.edge.quantized_decomposed.quantize_per_tensor.default,
+    exir_ops.edge.quantized_decomposed.quantize_per_tensor.tensor,
+    exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default,
+    exir_ops.edge.quantized_decomposed.dequantize_per_tensor.tensor,
+    exir_ops.edge.quantized_decomposed.dequantize_per_channel.default,
 ]
 
 SUPPORTS_DYNAMIC_SHAPE = [

backends/vulkan/vulkan_preprocess.py

@@ -13,6 +13,7 @@
     FuseBatchNormWithConvPass,
 )
 from executorch.backends.transforms.fuse_conv_with_clamp import FuseClampPass
+from executorch.backends.transforms.fuse_dequant_linear import FuseDequantLinearPass
 from executorch.backends.transforms.fuse_view_copy import FuseViewCopyTransform
 from executorch.backends.transforms.mean_to_sum_div import MeanToSumDiv
 from executorch.backends.transforms.remove_clone_ops import RemoveCloneOpsTransform
@@ -59,6 +60,7 @@ def preprocess(  # noqa: C901
         passes = [
             RemoveCloneOpsTransform(),
             AddmmToLinearTransform(),
+            FuseDequantLinearPass(),
             FuseViewCopyTransform(),
             FuseBatchNormWithConvPass(program),
             FuseClampPass(),