ipex-llm/docker/llm/serving/xpu/docker/ccl_torch.patch
Guancheng Fu af693425f1
Upgrade to vLLM 0.6.6 (#12796)
* init

* update engine init

* fix serving load_in_low_bit problem

* temp

* temp

* temp

* temp

* temp

* fix

* fixed

* done

* fix

* fix all arguments

* fix

* fix throughput script

* fix

* fix

* use official ipex-llm

* Fix readme

* fix

---------

Co-authored-by: hzjane <a1015616934@qq.com>
2025-02-12 16:47:51 +08:00

47 lines
1.9 KiB
Diff

diff --git a/setup.py b/setup.py
index 67016cb..305be5c 100644
--- a/setup.py
+++ b/setup.py
@@ -118,8 +118,8 @@ class BuildCMakeExt(BuildExtension):
if compute_backend == 'dpcpp':
runtime = 'dpcpp'
build_options['COMPUTE_BACKEND'] = compute_backend
- import intel_extension_for_pytorch
- build_options['CMAKE_PREFIX_PATH'] += ";" + intel_extension_for_pytorch.cmake_prefix_path
+ # import intel_extension_for_pytorch
+ # build_options['CMAKE_PREFIX_PATH'] += ";" + intel_extension_for_pytorch.cmake_prefix_path
if "DPCPP_GCC_INSTALL_DIR" in my_env:
exist_cflags = "CFLAGS" in my_env
cflags = ""
diff --git a/src/gpu/CMakeLists.txt b/src/gpu/CMakeLists.txt
index 5628f31..d1589c4 100644
--- a/src/gpu/CMakeLists.txt
+++ b/src/gpu/CMakeLists.txt
@@ -1,4 +1,4 @@
-find_package(IPEX REQUIRED)
+# find_package(IPEX REQUIRED)
set(CCL_DPCPP_SRCS dpcpp_ccl.cpp ze_exception.hpp allreduce.h sycl_misc.hpp runtime.hpp cxxopts.hpp)
@@ -9,7 +9,7 @@ add_library(oneccl_bindings_for_pytorch_xpu SHARED ${CCL_DPCPP_SRCS})
target_link_libraries(oneccl_bindings_for_pytorch_xpu PUBLIC ${DEPENDS_LIB})
target_link_libraries(oneccl_bindings_for_pytorch_xpu PUBLIC oneccl_bindings_for_pytorch)
-target_link_libraries(oneccl_bindings_for_pytorch_xpu PUBLIC intel-ext-pt-gpu)
+# target_link_libraries(oneccl_bindings_for_pytorch_xpu PUBLIC intel-ext-pt-gpu)
foreach(RPATH ${CMAKE_INSTALL_RPATH})
set_target_properties(oneccl_bindings_for_pytorch_xpu PROPERTIES LINK_FLAGS "-Wl,-rpath,${RPATH}")
diff --git a/src/gpu/dpcpp_ccl.cpp b/src/gpu/dpcpp_ccl.cpp
index 1631b85..0945031 100644
--- a/src/gpu/dpcpp_ccl.cpp
+++ b/src/gpu/dpcpp_ccl.cpp
@@ -32,7 +32,7 @@
#include <ATen/record_function.h>
#include <ProcessGroupCCL.hpp>
#include <dispatch_stub.h>
-#include <ipex.h>
+// #include <ipex.h>
#include <sycl/sycl.hpp>
//#include "allreduce.h"