Merge branch 'main' into encoder

CMakeLists.txt

@@ -22,6 +22,9 @@ cmake_minimum_required(VERSION 3.13 FATAL_ERROR)
 
 set(PROJECT_NAME "torch-xpu-ops")
 set(PROJECT_VERSION "2.3.0")
+# Avoid SYCL compiler error
+string(APPEND CMAKE_CXX_FLAGS " -Wno-error=comment")
+
 cmake_policy(SET CMP0048 NEW)
 project(${PROJECT_NAME} VERSION "${PROJECT_VERSION}" LANGUAGES C CXX)
 

test/xpu/skip_list_common.py

@@ -2688,16 +2688,8 @@
     "nn/test_pruning_xpu.py": None,
 
     "test_foreach_xpu.py": (
-        # CPU fallback fails. Implementation difference between CPU and CUDA. Expect success on CPU and expect fail on CUDA. When we use CPU fallback and align expected fail list with CUDA, these cases fail.
-        "test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_xpu_bool",
-        # AssertionError: RuntimeError not raised
-        # https://github.com/intel/torch-xpu-ops/issues/784 
-        "test_0dim_tensor_overload_exception_xpu",
         # RuntimeError: Tried to instantiate dummy base class CUDAGraph
-        "test_big_num_tensors__foreach_max_use_cuda_graph_True_xpu_float32",
-        "test_big_num_tensors__foreach_max_use_cuda_graph_True_xpu_float64",
-        "test_big_num_tensors__foreach_norm_use_cuda_graph_True_xpu_float32",
-        "test_big_num_tensors__foreach_norm_use_cuda_graph_True_xpu_float64",
+        "use_cuda_graph_True",
     ),
 
     "nn/test_convolution_xpu.py": (