diff --git a/build2cmake/src/templates/cpu/torch-binding.cmake b/build2cmake/src/templates/cpu/torch-binding.cmake
deleted file mode 100644
index 799fa0d3..00000000
--- a/build2cmake/src/templates/cpu/torch-binding.cmake
+++ /dev/null
@@ -1,13 +0,0 @@
-set(TORCH_{{name}}_SRC
-  {{ src|join(' ') }}
-)
-
-{% if includes %}
-# TODO: check if CLion support this:
-# https://youtrack.jetbrains.com/issue/CPP-16510/CLion-does-not-handle-per-file-include-directories
-set_source_files_properties(
-  {{'${TORCH_' + name + '_SRC}'}}
-  PROPERTIES INCLUDE_DIRECTORIES "{{ includes }}")
-{% endif %}
-
-list(APPEND SRC {{'"${TORCH_' + name + '_SRC}"'}})
diff --git a/build2cmake/src/templates/cpu/torch-extension.cmake b/build2cmake/src/templates/cpu/torch-extension.cmake
deleted file mode 100644
index 9d667e4a..00000000
--- a/build2cmake/src/templates/cpu/torch-extension.cmake
+++ /dev/null
@@ -1,9 +0,0 @@
-define_gpu_extension_target(
-  {{ ops_name }}
-  DESTINATION {{ ops_name }}
-  LANGUAGE ${GPU_LANG}
-  SOURCES ${SRC}
-  COMPILE_FLAGS ${GPU_FLAGS}
-  ARCHITECTURES ${GPU_ARCHES}
-  USE_SABI 3
-  WITH_SOABI)
diff --git a/build2cmake/src/templates/cuda/torch-binding.cmake b/build2cmake/src/templates/cuda/torch-binding.cmake
deleted file mode 100644
index 5c1c0002..00000000
--- a/build2cmake/src/templates/cuda/torch-binding.cmake
+++ /dev/null
@@ -1,16 +0,0 @@
-get_torch_gpu_compiler_flags(TORCH_GPU_FLAGS ${GPU_LANG})
-list(APPEND GPU_FLAGS ${TORCH_GPU_FLAGS})
-
-set(TORCH_{{name}}_SRC
-  {{ src|join(' ') }}
-)
-
-{% if includes %}
-# TODO: check if CLion support this:
-# https://youtrack.jetbrains.com/issue/CPP-16510/CLion-does-not-handle-per-file-include-directories
-set_source_files_properties(
-  {{'${TORCH_' + name + '_SRC}'}}
-  PROPERTIES INCLUDE_DIRECTORIES "{{ includes }}")
-{% endif %}
-
-list(APPEND SRC {{'"${TORCH_' + name + '_SRC}"'}})
diff --git a/build2cmake/src/templates/cuda/torch-extension.cmake b/build2cmake/src/templates/cuda/torch-extension.cmake
deleted file mode 100644
index 5b934e10..00000000
--- a/build2cmake/src/templates/cuda/torch-extension.cmake
+++ /dev/null
@@ -1,25 +0,0 @@
-define_gpu_extension_target(
-  {{ ops_name }}
-  DESTINATION {{ ops_name }}
-  LANGUAGE ${GPU_LANG}
-  SOURCES ${SRC}
-  COMPILE_FLAGS ${GPU_FLAGS}
-  ARCHITECTURES ${GPU_ARCHES}
-  #INCLUDE_DIRECTORIES ${CUTLASS_INCLUDE_DIR}
-  USE_SABI 3
-  WITH_SOABI)
-
-if( NOT MSVC)
-    target_link_options({{ ops_name }} PRIVATE -static-libstdc++)
-endif()
-
-{% if platform == 'windows' %}
-# These methods below should be included from preamble.cmake on windows platform.
-
-# Add kernels_install target for huggingface/kernels library layout
-add_kernels_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
-
-# Add local_install target for local development with get_local_kernel()
-add_local_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
-
-{% endif %}
diff --git a/build2cmake/src/templates/metal/torch-extension.cmake b/build2cmake/src/templates/metal/torch-extension.cmake
deleted file mode 100644
index 9e9fa7aa..00000000
--- a/build2cmake/src/templates/metal/torch-extension.cmake
+++ /dev/null
@@ -1,17 +0,0 @@
-# Include Metal shader compilation utilities
-include(${CMAKE_CURRENT_LIST_DIR}/cmake/compile-metal.cmake)
-
-define_gpu_extension_target(
-  {{ ops_name }}
-  DESTINATION {{ ops_name }}
-  LANGUAGE ${GPU_LANG}
-  SOURCES ${SRC}
-  COMPILE_FLAGS ${GPU_FLAGS}
-  ARCHITECTURES ${GPU_ARCHES}
-  USE_SABI 3
-  WITH_SOABI)
-
-# Compile Metal shaders if any were found
-if(ALL_METAL_SOURCES)
-  compile_metal_shaders({{ ops_name }} "${ALL_METAL_SOURCES}" "${METAL_INCLUDE_DIRS}")
-endif()
\ No newline at end of file
diff --git a/build2cmake/src/templates/metal/torch-binding.cmake b/build2cmake/src/templates/torch-binding.cmake
similarity index 73%
rename from build2cmake/src/templates/metal/torch-binding.cmake
rename to build2cmake/src/templates/torch-binding.cmake
index 79872f04..93e47325 100644
--- a/build2cmake/src/templates/metal/torch-binding.cmake
+++ b/build2cmake/src/templates/torch-binding.cmake
@@ -1,5 +1,7 @@
-#get_torch_gpu_compiler_flags(TORCH_GPU_FLAGS ${GPU_LANG})
-#list(APPEND GPU_FLAGS ${TORCH_GPU_FLAGS})
+if(GPU_LANG STREQUAL "CUDA")
+  get_torch_gpu_compiler_flags(TORCH_GPU_FLAGS ${GPU_LANG})
+  list(APPEND GPU_FLAGS ${TORCH_GPU_FLAGS})
+endif()
 
 set(TORCH_{{name}}_SRC
   {{ src|join(' ') }}
diff --git a/build2cmake/src/templates/torch-extension.cmake b/build2cmake/src/templates/torch-extension.cmake
new file mode 100644
index 00000000..2fdd220c
--- /dev/null
+++ b/build2cmake/src/templates/torch-extension.cmake
@@ -0,0 +1,39 @@
+# Include Metal shader compilation utilities if needed
+if(GPU_LANG STREQUAL "METAL")
+  include(${CMAKE_CURRENT_LIST_DIR}/cmake/compile-metal.cmake)
+endif()
+
+# Define the extension target with unified parameters
+define_gpu_extension_target(
+  {{ ops_name }}
+  DESTINATION {{ ops_name }}
+  LANGUAGE ${GPU_LANG}
+  SOURCES ${SRC}
+  COMPILE_FLAGS ${GPU_FLAGS}
+  ARCHITECTURES ${GPU_ARCHES}
+  USE_SABI 3
+  WITH_SOABI)
+
+if(NOT (MSVC OR GPU_LANG STREQUAL "SYCL"))
+  target_link_options({{ ops_name }} PRIVATE -static-libstdc++)
+endif()
+
+if(GPU_LANG STREQUAL "SYCL")
+  target_link_options({{ ops_name }} PRIVATE ${sycl_link_flags})
+  target_link_libraries({{ ops_name }} PRIVATE dnnl)
+endif()
+
+# Compile Metal shaders if any were found
+if(GPU_LANG STREQUAL "METAL")
+  if(ALL_METAL_SOURCES)
+    compile_metal_shaders({{ ops_name }} "${ALL_METAL_SOURCES}" "${METAL_INCLUDE_DIRS}")
+  endif()
+endif()
+
+{% if platform == 'windows' %}
+# Add kernels_install target for huggingface/kernels library layout
+add_kernels_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
+
+# Add local_install target for local development with get_local_kernel()
+add_local_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
+{% endif %}
diff --git a/build2cmake/src/templates/xpu/preamble.cmake b/build2cmake/src/templates/xpu/preamble.cmake
index aae15071..1687264a 100644
--- a/build2cmake/src/templates/xpu/preamble.cmake
+++ b/build2cmake/src/templates/xpu/preamble.cmake
@@ -79,7 +79,8 @@ add_compile_definitions(USE_XPU)
 # Set SYCL-specific flags
 # Set comprehensive SYCL compilation and linking flags
 set(sycl_link_flags "-fsycl;--offload-compress;-fsycl-targets=spir64_gen,spir64;-Xs;-device pvc,xe-lpg,ats-m150 -options ' -cl-intel-enable-auto-large-GRF-mode -cl-poison-unsupported-fp64-kernels -cl-intel-greater-than-4GB-buffer-required';")
-set(sycl_flags "-fsycl;-fhonor-nans;-fhonor-infinities;-fno-associative-math;-fno-approx-func;-fno-sycl-instrument-device-code;--offload-compress;-fsycl-targets=spir64_gen,spir64;")
+set(GPU_FLAGS "-fsycl;-fhonor-nans;-fhonor-infinities;-fno-associative-math;-fno-approx-func;-fno-sycl-instrument-device-code;--offload-compress;-fsycl-targets=spir64_gen,spir64;")
+set(GPU_ARCHES "")
 message(STATUS "Configuring for Intel XPU backend using SYCL")
 
 {% if platform == 'windows' %}
diff --git a/build2cmake/src/templates/xpu/torch-binding.cmake b/build2cmake/src/templates/xpu/torch-binding.cmake
deleted file mode 100644
index 8812c529..00000000
--- a/build2cmake/src/templates/xpu/torch-binding.cmake
+++ /dev/null
@@ -1,13 +0,0 @@
-set(TORCH_{{name}}_SRC
-  {{ src|join(' ') }}
-)
-
-{% if includes %}
-# TODO: check if CLion support this:
-# https://youtrack.jetbrains.com/issue/CPP-16510/CLion-does-not-handle-per-file-include-directories
-set_source_files_properties(
-  {{'${TORCH_' + name + '_SRC}'}}
-  PROPERTIES INCLUDE_DIRECTORIES "{{ includes }}")
-{% endif %}
-
-list(APPEND SRC {{'"${TORCH_' + name + '_SRC}"'}})
\ No newline at end of file
diff --git a/build2cmake/src/templates/xpu/torch-extension.cmake b/build2cmake/src/templates/xpu/torch-extension.cmake
deleted file mode 100644
index ae7a69aa..00000000
--- a/build2cmake/src/templates/xpu/torch-extension.cmake
+++ /dev/null
@@ -1,23 +0,0 @@
-define_gpu_extension_target(
-  {{ ops_name }}
-  DESTINATION {{ ops_name }}
-  LANGUAGE ${GPU_LANG}
-  SOURCES ${SRC}
-  COMPILE_FLAGS ${sycl_flags}
-  USE_SABI 3
-  WITH_SOABI)
-
-# Add XPU/SYCL specific linker flags
-target_link_options({{ ops_name }} PRIVATE ${sycl_link_flags})
-target_link_libraries({{ ops_name }} PRIVATE dnnl)
-
-{% if platform == 'windows' %}
-# These methods below should be included from preamble.cmake on windows platform.
-
-# Add kernels_install target for huggingface/kernels library layout
-add_kernels_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
-
-# Add local_install target for local development with get_local_kernel()
-add_local_install_target({{ ops_name }} "{{ name }}" "${BUILD_VARIANT_NAME}")
-
-{% endif %}
\ No newline at end of file
diff --git a/build2cmake/src/torch/common.rs b/build2cmake/src/torch/common.rs
index c5eb162d..08200340 100644
--- a/build2cmake/src/torch/common.rs
+++ b/build2cmake/src/torch/common.rs
@@ -1,8 +1,10 @@
+use std::io::Write;
+
 use eyre::{Context, Result};
 use itertools::Itertools;
 use minijinja::{context, Environment};
 
-use crate::config::{Backend, General};
+use crate::config::{Backend, General, Torch};
 use crate::metadata::Metadata;
 use crate::FileSet;
 
@@ -64,3 +66,49 @@ where
         .collect_vec()
         .join(";")
 }
+
+pub fn render_binding(
+    env: &Environment,
+    torch: &Torch,
+    name: &str,
+    write: &mut impl Write,
+) -> Result<()> {
+    env.get_template("torch-binding.cmake")
+        .wrap_err("Cannot get Torch binding template")?
+        .render_to_write(
+            context! {
+                includes => torch.include.as_ref().map(prefix_and_join_includes),
+                name => name,
+                src => torch.src
+            },
+            &mut *write,
+        )
+        .wrap_err("Cannot render Torch binding template")?;
+
+    write.write_all(b"\n")?;
+
+    Ok(())
+}
+
+pub fn render_extension(
+    env: &Environment,
+    name: &str,
+    ops_name: &str,
+    write: &mut impl Write,
+) -> Result<()> {
+    env.get_template("torch-extension.cmake")
+        .wrap_err("Cannot get Torch extension template")?
+        .render_to_write(
+            context! {
+                name => name,
+                ops_name => ops_name,
+                platform => std::env::consts::OS,
+            },
+            &mut *write,
+        )
+        .wrap_err("Cannot render Torch extension template")?;
+
+    write.write_all(b"\n")?;
+
+    Ok(())
+}
diff --git a/build2cmake/src/torch/cpu.rs b/build2cmake/src/torch/cpu.rs
index b93fa592..08c4e650 100644
--- a/build2cmake/src/torch/cpu.rs
+++ b/build2cmake/src/torch/cpu.rs
@@ -1,13 +1,13 @@
 use std::{io::Write, path::PathBuf};
 
 use eyre::{bail, Context, Result};
-use itertools::Itertools;
 use minijinja::{context, Environment};
 
 use crate::config::{Backend, Build, Torch};
 use crate::fileset::FileSet;
-use crate::torch::common::write_metadata;
-use crate::torch::common::write_pyproject_toml;
+use crate::torch::common::{
+    render_binding, render_extension, write_metadata, write_pyproject_toml,
+};
 use crate::torch::kernel::render_kernel_components;
 use crate::torch::kernel_ops_identifier;
 use crate::version::Version;
@@ -103,51 +103,6 @@ fn write_cmake(
     Ok(())
 }
 
-fn render_binding(
-    env: &Environment,
-    torch: &Torch,
-    name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("cpu/torch-binding.cmake")
-        .wrap_err("Cannot get Torch binding template")?
-        .render_to_write(
-            context! {
-                includes => torch.include.as_ref().map(prefix_and_join_includes),
-                name => name,
-                src => torch.src
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch binding template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
-pub fn render_extension(
-    env: &Environment,
-    name: &str,
-    ops_name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("cpu/torch-extension.cmake")
-        .wrap_err("Cannot get Torch extension template")?
-        .render_to_write(
-            context! {
-                name => name,
-                ops_name => ops_name,
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch extension template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 fn render_preamble(
     env: &Environment,
     name: &str,
@@ -234,15 +189,3 @@ fn write_torch_registration_macros(file_set: &mut FileSet) -> Result<()> {
 
     Ok(())
 }
-
-fn prefix_and_join_includes<S>(includes: impl AsRef<[S]>) -> String
-where
-    S: AsRef<str>,
-{
-    includes
-        .as_ref()
-        .iter()
-        .map(|include| format!("${{CMAKE_SOURCE_DIR}}/{}", include.as_ref()))
-        .collect_vec()
-        .join(";")
-}
diff --git a/build2cmake/src/torch/cuda.rs b/build2cmake/src/torch/cuda.rs
index 970f5bc3..e7c5acdc 100644
--- a/build2cmake/src/torch/cuda.rs
+++ b/build2cmake/src/torch/cuda.rs
@@ -7,9 +7,9 @@ use eyre::{bail, Context, Result};
 use minijinja::{context, Environment};
 
 use crate::config::{Backend, Build, Dependency, Torch};
-use crate::torch::common::prefix_and_join_includes;
-use crate::torch::common::write_metadata;
-use crate::torch::common::write_pyproject_toml;
+use crate::torch::common::{
+    render_binding, render_extension, write_metadata, write_pyproject_toml,
+};
 use crate::torch::kernel::render_kernel_components;
 use crate::torch::kernel_ops_identifier;
 use crate::version::Version;
@@ -189,29 +189,6 @@ fn write_cmake(
     Ok(())
 }
 
-pub fn render_binding(
-    env: &Environment,
-    torch: &Torch,
-    name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("cuda/torch-binding.cmake")
-        .wrap_err("Cannot get Torch binding template")?
-        .render_to_write(
-            context! {
-                includes => torch.include.as_ref().map(prefix_and_join_includes),
-                name => name,
-                src => torch.src
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch binding template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 fn render_deps(
     env: &Environment,
     backend: Backend,
@@ -307,29 +284,6 @@ fn render_deps(
     Ok(())
 }
 
-pub fn render_extension(
-    env: &Environment,
-    name: &str,
-    ops_name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("cuda/torch-extension.cmake")
-        .wrap_err("Cannot get Torch extension template")?
-        .render_to_write(
-            context! {
-                name => name,
-                ops_name => ops_name,
-                platform => std::env::consts::OS,
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch extension template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 pub fn render_preamble(
     env: &Environment,
     name: &str,
diff --git a/build2cmake/src/torch/metal.rs b/build2cmake/src/torch/metal.rs
index 93089fe9..60e47a5f 100644
--- a/build2cmake/src/torch/metal.rs
+++ b/build2cmake/src/torch/metal.rs
@@ -1,13 +1,13 @@
 use std::{io::Write, path::PathBuf};
 
 use eyre::{bail, Context, Result};
-use itertools::Itertools;
 use minijinja::{context, Environment};
 
 use crate::config::{Backend, Build, Torch};
 use crate::fileset::FileSet;
-use crate::torch::common::write_metadata;
-use crate::torch::common::write_pyproject_toml;
+use crate::torch::common::{
+    render_binding, render_extension, write_metadata, write_pyproject_toml,
+};
 use crate::torch::kernel::render_kernel_components;
 use crate::torch::kernel_ops_identifier;
 use crate::version::Version;
@@ -119,51 +119,6 @@ fn write_cmake(
     Ok(())
 }
 
-fn render_binding(
-    env: &Environment,
-    torch: &Torch,
-    name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("metal/torch-binding.cmake")
-        .wrap_err("Cannot get Torch binding template")?
-        .render_to_write(
-            context! {
-                includes => torch.include.as_ref().map(prefix_and_join_includes),
-                name => name,
-                src => torch.src
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch binding template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
-pub fn render_extension(
-    env: &Environment,
-    name: &str,
-    ops_name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("metal/torch-extension.cmake")
-        .wrap_err("Cannot get Torch extension template")?
-        .render_to_write(
-            context! {
-                name => name,
-                ops_name => ops_name,
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch extension template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 fn render_preamble(
     env: &Environment,
     name: &str,
@@ -250,15 +205,3 @@ fn write_torch_registration_macros(file_set: &mut FileSet) -> Result<()> {
 
     Ok(())
 }
-
-fn prefix_and_join_includes<S>(includes: impl AsRef<[S]>) -> String
-where
-    S: AsRef<str>,
-{
-    includes
-        .as_ref()
-        .iter()
-        .map(|include| format!("${{CMAKE_SOURCE_DIR}}/{}", include.as_ref()))
-        .collect_vec()
-        .join(";")
-}
diff --git a/build2cmake/src/torch/xpu.rs b/build2cmake/src/torch/xpu.rs
index 03738a68..3a14b65d 100644
--- a/build2cmake/src/torch/xpu.rs
+++ b/build2cmake/src/torch/xpu.rs
@@ -3,12 +3,12 @@ use std::io::Write;
 use std::path::PathBuf;
 
 use eyre::{bail, Context, Result};
-use itertools::Itertools;
 use minijinja::{context, Environment};
 
 use crate::config::{Backend, Build, Dependency, Torch};
-use crate::torch::common::write_metadata;
-use crate::torch::common::write_pyproject_toml;
+use crate::torch::common::{
+    render_binding, render_extension, write_metadata, write_pyproject_toml,
+};
 use crate::torch::kernel::render_kernel_components;
 use crate::torch::kernel_ops_identifier;
 use crate::version::Version;
@@ -176,29 +176,6 @@ fn write_cmake(
     Ok(())
 }
 
-fn render_binding(
-    env: &Environment,
-    torch: &Torch,
-    name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("xpu/torch-binding.cmake")
-        .wrap_err("Cannot get Torch binding template")?
-        .render_to_write(
-            context! {
-                includes => torch.include.as_ref().map(prefix_and_join_includes),
-                name => name,
-                src => torch.src
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch binding template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 fn render_deps(
     env: &Environment,
     backend: Backend,
@@ -233,29 +210,6 @@ fn render_deps(
     Ok(())
 }
 
-pub fn render_extension(
-    env: &Environment,
-    name: &str,
-    ops_name: &str,
-    write: &mut impl Write,
-) -> Result<()> {
-    env.get_template("xpu/torch-extension.cmake")
-        .wrap_err("Cannot get Torch extension template")?
-        .render_to_write(
-            context! {
-                name => name,
-                ops_name => ops_name,
-                platform => std::env::consts::OS
-            },
-            &mut *write,
-        )
-        .wrap_err("Cannot render Torch extension template")?;
-
-    write.write_all(b"\n")?;
-
-    Ok(())
-}
-
 pub fn render_preamble(
     env: &Environment,
     name: &str,
@@ -280,15 +234,3 @@ pub fn render_preamble(
 
     Ok(())
 }
-
-fn prefix_and_join_includes<S>(includes: impl AsRef<[S]>) -> String
-where
-    S: AsRef<str>,
-{
-    includes
-        .as_ref()
-        .iter()
-        .map(|include| format!("${{CMAKE_SOURCE_DIR}}/{}", include.as_ref()))
-        .collect_vec()
-        .join(";")
-}