~ruther/guix-local

720d2b57ebf1222cb998acaa46a91f5ee08f2041 — Ayan Das 5 months ago 04eeea1 
gnu: python-pytorch: Update to 2.9.0.

* gnu/packages/machine-learning.scm (python-pytorch): Update to 2.9.0.
  [source]: Update hash.
  [arguments]: Add 'skip-pip-redirect' phase to avoid pip redirect in
  setup.py.  Extend concurrentqueue include fix to cover nativert.
  [native-inputs]: Add python-setuptools.

* gnu/packages/patches/python-pytorch-system-libraries.patch: Update for 2.9.0.

Change-Id: I26606949da3c36fe8d7b6933ec54cb2c347a1bb9
Signed-off-by: Sharlatan Hellseher <sharlatanus@gmail.com>
patch browse .tar.gz 
2 files changed, 99 insertions(+), 101 deletions(-)

M gnu/packages/machine-learning.scm
M gnu/packages/patches/python-pytorch-system-libraries.patch

M gnu/packages/machine-learning.scm => gnu/packages/machine-learning.scm +12 -4
@@ 4308,7 4308,7 @@ PyTorch.")
        (base32
         "0hdpkhcjry22fjx2zg2r48v7f4ljrclzj0li2pgk76kvyblfbyvm"))))))

(define %python-pytorch-version "2.8.0")
(define %python-pytorch-version "2.9.0")

(define %python-pytorch-src
  (origin


@@ 4319,7 4319,7 @@ PyTorch.")
    (file-name (git-file-name "python-pytorch" %python-pytorch-version))
    (sha256
     (base32
      "0am8mx0mq3hqsk1g99a04a4fdf865g93568qr1f247pl11r2jldl"))
      "005gj27qikkgbibbk00z8xs9a8xms2fxapm53inp31zxm4853myh"))
    (patches (search-patches "python-pytorch-system-libraries.patch"
                             "python-pytorch-runpath.patch"
                             "python-pytorch-without-kineto.patch"


@@ 4476,8 4476,14 @@ PyTorch.")
          ;; the 'sanity-check phase to fail.
          (add-after 'unpack 'remove-fr-trace-script
            (lambda _
             (substitute* "setup.py"
               (("entry_points\\[\"console_scripts\"\\]\\.append\\(") "("))))
          (add-after 'remove-fr-trace-script 'skip-pip-redirect
            (lambda _
              ;; Keep using setup.py directly instead of invoking pip.
              (substitute* "setup.py"
                (("entry_points\\[\"console_scripts\"\\]\\.append\\(") "("))))
                (("if arg == \"install\":")
                 "if False and arg == \"install\":"))))
          (add-before 'build 'use-system-libraries
            (lambda _
              (for-each


@@ 4494,7 4500,8 @@ PyTorch.")

              ;; Fix moodycamel/concurrentqueue includes for system package
              (substitute* '("c10/util/Semaphore.h"
                             "c10/test/util/Semaphore_test.cpp")
                             "c10/test/util/Semaphore_test.cpp"
                             "torch/nativert/executor/ParallelGraphExecutor.cpp")
                (("<moodycamel/concurrentqueue\\.h>") "<concurrentqueue.h>")
                (("<moodycamel/lightweightsemaphore\\.h>") "<lightweightsemaphore.h>"))



@@ 4655,6 4662,7 @@ PyTorch.")
           python-pytest-shard
           python-pytest-xdist
           python-hypothesis
           python-setuptools
           python-types-dataclasses
           shaderc
           valgrind/pinned))


M gnu/packages/patches/python-pytorch-system-libraries.patch => gnu/packages/patches/python-pytorch-system-libraries.patch +87 -97
@@ 3,11 3,12 @@ libraries not supported or working only by specifying USE_SYSTEM_LIBS.  This
includes using the clog, cpuinfo, fbgemm, foxi, fp16, fxdiv, googletest,
httlib, ideep, miniz, nnpack, oneapi-dnnl, pocketfft, pthreadpool,
qnnpack-pytorch, tensorpipe, valgrind and xnnpack packages.

diff --git i/aten/src/ATen/CMakeLists.txt w/aten/src/ATen/CMakeLists.txt
index c9cfd74b501..590c5d61d17 100644
index 6c095680733..646d342acc6 100644
--- i/aten/src/ATen/CMakeLists.txt
+++ w/aten/src/ATen/CMakeLists.txt
@@ -506,9 +506,9 @@ if(AT_NNPACK_ENABLED)
@@ -600,9 +600,9 @@ if(AT_NNPACK_ENABLED)
   list(APPEND ATen_CPU_DEPENDENCY_LIBS nnpack) # cpuinfo is added below
 endif()
 


@@ 21,7 22,7 @@ index c9cfd74b501..590c5d61d17 100644
 if(USE_MKLDNN_ACL)
     list(APPEND ATen_CPU_INCLUDE ${ACL_INCLUDE_DIRS})
diff --git i/caffe2/CMakeLists.txt w/caffe2/CMakeLists.txt
index d1ce24e37fe..43e60deba86 100644
index 6ab41b6c847..f46610ec4fd 100644
--- i/caffe2/CMakeLists.txt
+++ w/caffe2/CMakeLists.txt
@@ -91,9 +91,6 @@ if(NOT MSVC AND USE_XNNPACK)


@@ 34,7 35,7 @@ index d1ce24e37fe..43e60deba86 100644
   endif()
 endif()
 
@@ -1195,7 +1192,6 @@ if(USE_XPU)
@@ -1205,7 +1202,6 @@ if(USE_XPU)
 endif()
 
 if(NOT MSVC AND USE_XNNPACK)


@@ 42,7 43,7 @@ index d1ce24e37fe..43e60deba86 100644
 endif()
 
 # ==========================================================
@@ -1315,8 +1311,8 @@ endif()
@@ -1325,8 +1321,8 @@ endif()
 target_include_directories(torch_cpu PRIVATE
   ${TORCH_ROOT}/third_party/cpp-httplib)
 


@@ 53,7 54,7 @@ index d1ce24e37fe..43e60deba86 100644
 
 install(DIRECTORY
   "${TORCH_SRC_DIR}/csrc"
@@ -1542,6 +1538,7 @@ target_link_libraries(torch_cpu PUBLIC c10)
@@ -1548,6 +1544,7 @@ target_link_libraries(torch_cpu PUBLIC c10)
 target_link_libraries(torch_cpu PUBLIC ${Caffe2_PUBLIC_DEPENDENCY_LIBS})
 target_link_libraries(torch_cpu PRIVATE ${Caffe2_DEPENDENCY_LIBS})
 target_link_libraries(torch_cpu PRIVATE ${Caffe2_DEPENDENCY_WHOLE_LINK_LIBS})


@@ 61,7 62,7 @@ index d1ce24e37fe..43e60deba86 100644
 if(USE_MPI)
   target_link_libraries(torch_cpu PRIVATE MPI::MPI_CXX)
 endif()
@@ -1793,7 +1790,7 @@ if(BUILD_MOBILE_TEST)
@@ -1815,7 +1812,7 @@ if(BUILD_MOBILE_TEST)
   foreach(test_src ${ATen_MOBILE_TEST_SRCS})
     get_filename_component(test_name ${test_src} NAME_WE)
     add_executable(${test_name} "${test_src}")


@@ 70,7 71,7 @@ index d1ce24e37fe..43e60deba86 100644
     target_include_directories(${test_name} PRIVATE $<INSTALL_INTERFACE:include>)
     target_include_directories(${test_name} PRIVATE $<BUILD_INTERFACE:${CMAKE_BINARY_DIR}/include>)
     target_include_directories(${test_name} PRIVATE ${ATen_CPU_INCLUDE})
@@ -1814,7 +1811,7 @@ if(BUILD_TEST)
@@ -1836,7 +1833,7 @@ if(BUILD_TEST)
         if(NOT MSVC)
           add_executable(${test_name}_${CPU_CAPABILITY} "${test_src}" ../aten/src/ATen/native/quantized/AffineQuantizerBase.cpp)
           # TODO: Get rid of c10 dependency (which is only needed for the implementation of AT_ERROR)


@@ 79,7 80,7 @@ index d1ce24e37fe..43e60deba86 100644
           if(USE_FBGEMM)
             target_link_libraries(${test_name}_${CPU_CAPABILITY} fbgemm)
           endif()
@@ -1828,7 +1825,7 @@ if(BUILD_TEST)
@@ -1850,7 +1847,7 @@ if(BUILD_TEST)
           endif()
         else()
           add_executable(${test_name}_${CPU_CAPABILITY} "${test_src}")


@@ 88,7 89,7 @@ index d1ce24e37fe..43e60deba86 100644
         endif()
         target_include_directories(${test_name}_${CPU_CAPABILITY} PRIVATE $<INSTALL_INTERFACE:include>)
         target_include_directories(${test_name}_${CPU_CAPABILITY} PRIVATE $<BUILD_INTERFACE:${CMAKE_BINARY_DIR}/include>)
@@ -1869,7 +1866,7 @@ if(BUILD_TEST)
@@ -1891,7 +1888,7 @@ if(BUILD_TEST)
       add_executable(${test_name} "${test_src}")
       find_library(metal NAMES Metal)
       find_library(foundation NAMES Foundation)


@@ 97,7 98,7 @@ index d1ce24e37fe..43e60deba86 100644
       target_include_directories(${test_name} PRIVATE $<INSTALL_INTERFACE:include>)
       target_include_directories(${test_name} PRIVATE $<BUILD_INTERFACE:${CMAKE_BINARY_DIR}/include>)
       target_include_directories(${test_name} PRIVATE ${Caffe2_CPU_INCLUDE})
@@ -1889,7 +1886,7 @@ if(BUILD_TEST)
@@ -1911,7 +1908,7 @@ if(BUILD_TEST)
     foreach(test_src ${Caffe2_GPU_TEST_SRCS})
       get_filename_component(test_name ${test_src} NAME_WE)
       add_executable(${test_name} "${test_src}")


@@ 106,7 107,7 @@ index d1ce24e37fe..43e60deba86 100644
       if(USE_CUDNN AND ${test_name} MATCHES "cudnn")
         target_link_libraries(${test_name} torch::cudnn)
       endif()
@@ -1911,7 +1908,7 @@ if(BUILD_TEST)
@@ -1933,7 +1930,7 @@ if(BUILD_TEST)
     foreach(test_src ${Caffe2_XPU_TEST_SRCS})
       get_filename_component(test_name ${test_src} NAME_WE)
       add_executable(${test_name} "${test_src}")


@@ 115,7 116,7 @@ index d1ce24e37fe..43e60deba86 100644
       target_include_directories(${test_name} PRIVATE $<INSTALL_INTERFACE:include>)
       target_include_directories(${test_name} PRIVATE ${Caffe2_CPU_INCLUDE})
       add_test(NAME ${test_name} COMMAND $<TARGET_FILE:${test_name}>)
@@ -1926,7 +1923,7 @@ if(BUILD_TEST)
@@ -1948,7 +1945,7 @@ if(BUILD_TEST)
     foreach(test_src ${Caffe2_VULKAN_TEST_SRCS})
       get_filename_component(test_name ${test_src} NAME_WE)
       add_executable(${test_name} "${test_src}")


@@ 124,7 125,7 @@ index d1ce24e37fe..43e60deba86 100644
       target_include_directories(${test_name} PRIVATE $<INSTALL_INTERFACE:include>)
       target_include_directories(${test_name} PRIVATE ${Caffe2_CPU_INCLUDE})
       add_test(NAME ${test_name} COMMAND $<TARGET_FILE:${test_name}>)
@@ -1958,7 +1955,7 @@ if(BUILD_TEST)
@@ -1980,7 +1977,7 @@ if(BUILD_TEST)
       if(test_src MATCHES "^.*\.hip$")
         set_target_properties(${test_name} PROPERTIES LINKER_LANGUAGE CXX)
       endif()


@@ 146,10 147,10 @@ index ebbff0f292a..dcded25908e 100644
   ${CMAKE_CURRENT_SOURCE_DIR}/istream_adapter.cc
   ${CMAKE_CURRENT_SOURCE_DIR}/file_adapter.cc
diff --git i/cmake/Dependencies.cmake w/cmake/Dependencies.cmake
index a93386c27f8..43a058b23f9 100644
index ef5c2fd4e97..d6400cb616f 100644
--- i/cmake/Dependencies.cmake
+++ w/cmake/Dependencies.cmake
@@ -276,7 +276,7 @@ endif()
@@ -299,7 +299,7 @@ endif()
 # --- [ PocketFFT
 set(AT_POCKETFFT_ENABLED 0)
 if(NOT AT_MKL_ENABLED)


@@ 158,7 159,7 @@ index a93386c27f8..43a058b23f9 100644
   if(NOT EXISTS "${POCKETFFT_INCLUDE_DIR}")
     message(FATAL_ERROR "pocketfft directory not found, expected ${POCKETFFT_INCLUDE_DIR}")
   elseif(NOT EXISTS "${POCKETFFT_INCLUDE_DIR}/pocketfft_hdronly.h")
@@ -460,15 +460,6 @@ if(USE_PYTORCH_QNNPACK)
@@ -483,15 +483,6 @@ if(USE_PYTORCH_QNNPACK)
       set(PYTORCH_QNNPACK_BUILD_TESTS OFF CACHE BOOL "")
       set(PYTORCH_QNNPACK_BUILD_BENCHMARKS OFF CACHE BOOL "")
       set(PYTORCH_QNNPACK_LIBRARY_TYPE "static" CACHE STRING "")


@@ 174,7 175,7 @@ index a93386c27f8..43a058b23f9 100644
     endif()
 
     list(APPEND Caffe2_DEPENDENCY_LIBS pytorch_qnnpack)
@@ -558,16 +549,15 @@ if(USE_XNNPACK AND NOT USE_SYSTEM_XNNPACK)
@@ -581,16 +572,15 @@ if(USE_XNNPACK AND NOT USE_SYSTEM_XNNPACK)
   list(APPEND Caffe2_DEPENDENCY_LIBS XNNPACK microkernels-prod)
 elseif(NOT TARGET XNNPACK AND USE_SYSTEM_XNNPACK)
   add_library(XNNPACK SHARED IMPORTED)


@@ 184,7 185,7 @@ index a93386c27f8..43a058b23f9 100644
-  find_library(microkernels-prod_LIBRARY microkernels-prod)
   set_property(TARGET XNNPACK PROPERTY IMPORTED_LOCATION "${XNNPACK_LIBRARY}")
-  set_property(TARGET microkernels-prod PROPERTY IMPORTED_LOCATION "${microkernels-prod_LIBRARY}")
-  if(NOT XNNPACK_LIBRARY or NOT microkernels-prod_LIBRARY)
-  if(NOT XNNPACK_LIBRARY OR NOT microkernels-prod_LIBRARY)
+  set_property(TARGET microkernels-prod PROPERTY INTERFACE_LINK_LIBRARIES XNNPACK)
+  if(NOT XNNPACK_LIBRARY)
     message(FATAL_ERROR "Cannot find XNNPACK")


@@ 195,7 196,7 @@ index a93386c27f8..43a058b23f9 100644
 endif()
 
 # ---[ Vulkan deps
@@ -621,9 +611,6 @@ if(BUILD_TEST OR BUILD_MOBILE_BENCHMARK OR BUILD_MOBILE_TEST)
@@ -644,9 +634,6 @@ if(BUILD_TEST OR BUILD_MOBILE_BENCHMARK OR BUILD_MOBILE_TEST)
   set(INSTALL_GTEST OFF CACHE BOOL "Install gtest." FORCE)
   set(BUILD_GMOCK ON CACHE BOOL "Build gmock." FORCE)
 


@@ 205,40 206,36 @@ index a93386c27f8..43a058b23f9 100644
 
   # We will not need to test benchmark lib itself.
   set(BENCHMARK_ENABLE_TESTING OFF CACHE BOOL "Disable benchmark testing as we don't need it.")
@@ -676,30 +663,14 @@ if(USE_FBGEMM)
     if(USE_ASAN)
       set(USE_SANITIZER "address,undefined" CACHE STRING "-fsanitize options for FBGEMM")
     endif()
@@ -678,16 +665,22 @@ if(USE_FBGEMM)
     set(FBGEMM_BUILD_TESTS OFF CACHE BOOL "")
     set(FBGEMM_BUILD_BENCHMARKS OFF CACHE BOOL "")
     set(FBGEMM_LIBRARY_TYPE "static" CACHE STRING "")
-    add_subdirectory("${FBGEMM_SOURCE_DIR}")
-    set_property(TARGET fbgemm_generic PROPERTY POSITION_INDEPENDENT_CODE ON)
-    set_property(TARGET fbgemm_avx2 PROPERTY POSITION_INDEPENDENT_CODE ON)
-    set_property(TARGET fbgemm_avx512 PROPERTY POSITION_INDEPENDENT_CODE ON)
-    set_property(TARGET fbgemm PROPERTY POSITION_INDEPENDENT_CODE ON)
-
-    # Disabling autovec in fbgemm due to large library size causing symbol relocation issues, which is only allowed in static builds.
-    # Long-term solution involves modularizing fbgemm targets.
-    target_compile_definitions(fbgemm_generic PUBLIC DISABLE_FBGEMM_AUTOVEC)
-    target_compile_definitions(fbgemm_avx2 PUBLIC DISABLE_FBGEMM_AUTOVEC)
-    target_compile_definitions(fbgemm_avx512 PUBLIC DISABLE_FBGEMM_AUTOVEC)
-
-    if("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang" AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 13.0.0)
-      # See https://github.com/pytorch/pytorch/issues/74352
-      target_compile_options_if_supported(asmjit -Wno-deprecated-copy)
-      target_compile_options_if_supported(asmjit -Wno-unused-but-set-variable)
-    endif()
     if(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
       target_compile_options_if_supported(asmjit -Wno-extra-semi)
       target_compile_options_if_supported(fbgemm -Wno-extra-semi)
-      target_compile_options_if_supported(asmjit -Wno-extra-semi)
-      target_compile_options_if_supported(fbgemm -Wno-extra-semi)
+      if(TARGET asmjit)
+        target_compile_options_if_supported(asmjit -Wno-extra-semi)
+      endif()
+      if(TARGET fbgemm)
+        target_compile_options_if_supported(fbgemm -Wno-extra-semi)
+      endif()
+    endif()
+    if(TARGET asmjit)
+      target_compile_options_if_supported(asmjit -Wno-unused-but-set-variable)
+      target_compile_options_if_supported(asmjit -Wno-unused-variable)
     endif()
-    target_compile_options_if_supported(asmjit -Wno-unused-but-set-variable)
-    target_compile_options_if_supported(asmjit -Wno-unused-variable)
   endif()
   if(USE_FBGEMM)
-    target_compile_definitions(fbgemm PUBLIC DISABLE_FBGEMM_AUTOVEC)
+    # Don't set compile definitions on system fbgemm
+    # target_compile_definitions(fbgemm PUBLIC DISABLE_FBGEMM_AUTOVEC)
     list(APPEND Caffe2_DEPENDENCY_LIBS fbgemm)
   endif()
 endif()
@@ -773,7 +744,7 @@ if(NOT TARGET fp16 AND NOT USE_SYSTEM_FP16)
@@ -758,7 +751,7 @@ if(NOT TARGET fp16 AND NOT USE_SYSTEM_FP16)
       "${CONFU_DEPENDENCIES_BINARY_DIR}/FP16")
   endif()
 elseif(NOT TARGET fp16 AND USE_SYSTEM_FP16)


@@ 247,15 244,21 @@ index a93386c27f8..43a058b23f9 100644
   set_target_properties(fp16 PROPERTIES LINKER_LANGUAGE C)
 endif()
 list(APPEND Caffe2_DEPENDENCY_LIBS fp16)
@@ -1156,7 +1127,6 @@ if(USE_DISTRIBUTED AND USE_TENSORPIPE)
       message(WARNING "Archived TensorPipe forces CMake compatibility mode")
       set(CMAKE_POLICY_VERSION_MINIMUM 3.5)
     endif()
@@ -1141,10 +1134,11 @@ if(USE_DISTRIBUTED AND USE_TENSORPIPE)
 
     # Tensorpipe uses cuda_add_library
     torch_update_find_cuda_flags()
-    add_subdirectory(${PROJECT_SOURCE_DIR}/third_party/tensorpipe)
     # Suppress warning to unblock libnop compilation by clang-17
     # See https://github.com/pytorch/pytorch/issues/151316
     target_compile_options_if_supported(tensorpipe -Wno-missing-template-arg-list-after-template-kw)
@@ -1331,7 +1301,7 @@ if(CAFFE2_CMAKE_BUILDING_WITH_MAIN_REPO AND NOT INTERN_DISABLE_ONNX)
-    target_compile_options_if_supported(tensorpipe -Wno-missing-template-arg-list-after-template-kw)
+    if(TARGET tensorpipe)
+      target_compile_options_if_supported(tensorpipe -Wno-missing-template-arg-list-after-template-kw)
+    endif()
 
     list(APPEND Caffe2_DEPENDENCY_LIBS tensorpipe)
     list(APPEND Caffe2_DEPENDENCY_LIBS nlohmann)
@@ -1320,7 +1314,7 @@ if(CAFFE2_CMAKE_BUILDING_WITH_MAIN_REPO AND NOT INTERN_DISABLE_ONNX)
     endif()
     set_property(TARGET onnx_proto PROPERTY IMPORTED_LOCATION ${ONNX_PROTO_LIBRARY})
     message("-- Found onnx: ${ONNX_LIBRARY} ${ONNX_PROTO_LIBRARY}")


@@ 264,7 267,7 @@ index a93386c27f8..43a058b23f9 100644
   endif()
   # Recover the build shared libs option.
   set(BUILD_SHARED_LIBS ${TEMP_BUILD_SHARED_LIBS})
@@ -1491,9 +1461,8 @@ if(NOT INTERN_BUILD_MOBILE)
@@ -1480,9 +1474,8 @@ if(NOT INTERN_BUILD_MOBILE)
   endif()
   if(USE_MKLDNN)
     include(${CMAKE_CURRENT_LIST_DIR}/public/mkldnn.cmake)


@@ 275,7 278,7 @@ index a93386c27f8..43a058b23f9 100644
     else()
       message(WARNING "MKLDNN could not be found.")
       caffe2_update_option(USE_MKLDNN OFF)
@@ -1557,7 +1526,7 @@ endif()
@@ -1546,7 +1539,7 @@ endif()
 #
 set(TEMP_BUILD_SHARED_LIBS ${BUILD_SHARED_LIBS})
 set(BUILD_SHARED_LIBS OFF CACHE BOOL "Build shared libs" FORCE)


@@ 284,7 287,7 @@ index a93386c27f8..43a058b23f9 100644
 
 # Disable compiler feature checks for `fmt`.
 #
@@ -1566,7 +1535,6 @@ add_subdirectory(${PROJECT_SOURCE_DIR}/third_party/fmt)
@@ -1555,7 +1548,6 @@ add_subdirectory(${PROJECT_SOURCE_DIR}/third_party/fmt)
 # CMAKE_CXX_FLAGS in ways that break feature checks. Since we already know
 # `fmt` is compatible with a superset of the compilers that PyTorch is, it
 # shouldn't be too bad to just disable the checks.


@@ 292,7 295,7 @@ index a93386c27f8..43a058b23f9 100644
 
 list(APPEND Caffe2_DEPENDENCY_LIBS fmt::fmt-header-only)
 set(BUILD_SHARED_LIBS ${TEMP_BUILD_SHARED_LIBS} CACHE BOOL "Build shared libs" FORCE)
@@ -1732,4 +1700,5 @@ include_directories(nlohmann SYSTEM INTERFACE ${PROJECT_SOURCE_DIR}/third_party/
@@ -1721,4 +1713,5 @@ include_directories(nlohmann SYSTEM INTERFACE ${PROJECT_SOURCE_DIR}/third_party/
 
 # Include moodycamel
 add_library(moodycamel INTERFACE IMPORTED)


@@ 341,28 344,28 @@ index 87935625f9b..9f8fa3df81f 100644
-  ${MKLDNN_LIBRARIES})
+  DNNL::dnnl)
diff --git i/setup.py w/setup.py
index b4ebc92f592..c3a9c4ab441 100644
index 11ca48482a7..0bc17caef68 100644
--- i/setup.py
+++ w/setup.py
@@ -556,13 +556,9 @@ def build_deps():
@@ -1025,13 +1025,9 @@ def build_deps() -> None:
     # Windows has very poor support for them.
     sym_files = [
         "tools/shared/_utils_internal.py",
-        "torch/utils/benchmark/utils/valgrind_wrapper/callgrind.h",
-        "torch/utils/benchmark/utils/valgrind_wrapper/valgrind.h",
         CWD / "tools/shared/_utils_internal.py",
-        CWD / "torch/utils/benchmark/utils/valgrind_wrapper/callgrind.h",
-        CWD / "torch/utils/benchmark/utils/valgrind_wrapper/valgrind.h",
     ]
     orig_files = [
         "torch/_utils_internal.py",
-        "third_party/valgrind-headers/callgrind.h",
-        "third_party/valgrind-headers/valgrind.h",
         CWD / "torch/_utils_internal.py",
-        CWD / "third_party/valgrind-headers/callgrind.h",
-        CWD / "third_party/valgrind-headers/valgrind.h",
     ]
     for sym_file, orig_file in zip(sym_files, orig_files):
         same = False
diff --git i/test/cpp/aoti_abi_check/CMakeLists.txt w/test/cpp/aoti_abi_check/CMakeLists.txt
index b317e040015..a42fd79da5a 100644
index 6898e406fb3..5400c159dac 100644
--- i/test/cpp/aoti_abi_check/CMakeLists.txt
+++ w/test/cpp/aoti_abi_check/CMakeLists.txt
@@ -20,7 +20,7 @@ target_compile_definitions(test_aoti_abi_check PRIVATE USE_GTEST)
@@ -28,7 +28,7 @@ target_compile_definitions(test_aoti_abi_check PRIVATE USE_GTEST)
 
 # WARNING: DO NOT LINK torch!!!
 # The purpose is to check if the used aten/c10 headers are writtern in a header-only way


@@ 370,12 373,21 @@ index b317e040015..a42fd79da5a 100644
+target_link_libraries(test_aoti_abi_check PRIVATE gtest_main gtest)
 target_include_directories(test_aoti_abi_check PRIVATE ${ATen_CPU_INCLUDE})
 
 if(INSTALL_TEST)
 foreach(test_src ${AOTI_ABI_CHECK_VEC_TEST_SRCS})
@@ -39,7 +39,7 @@ foreach(test_src ${AOTI_ABI_CHECK_VEC_TEST_SRCS})
     separate_arguments(FLAGS UNIX_COMMAND "${FLAGS}")
     add_executable(${test_name}_${CPU_CAPABILITY} "${test_src}")
 
-    target_link_libraries(${test_name}_${CPU_CAPABILITY} PRIVATE gtest_main)
+    target_link_libraries(${test_name}_${CPU_CAPABILITY} PRIVATE gtest_main gtest)
     target_include_directories(${test_name}_${CPU_CAPABILITY} PRIVATE ${ATen_CPU_INCLUDE})
 
     # Define CPU_CAPABILITY and CPU_CAPABILITY_XXX macros for conditional compilation
diff --git i/test/cpp/aoti_inference/CMakeLists.txt w/test/cpp/aoti_inference/CMakeLists.txt
index 5ac32ef3b91..a5dc1547899 100644
index cd87ba6c505..fc68301451a 100644
--- i/test/cpp/aoti_inference/CMakeLists.txt
+++ w/test/cpp/aoti_inference/CMakeLists.txt
@@ -55,7 +55,7 @@ add_custom_command(
@@ -56,7 +56,7 @@ add_custom_command(
 
 target_link_libraries(test_aoti_inference PRIVATE
   torch


@@ 420,8 432,8 @@ index 285a5dd2a73..51e76302265 100644
   if(USE_GLOO AND USE_C10D_GLOO)
-    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main INSTALL_TEST ${INSTALL_TEST})
-    c10d_add_test(ProcessGroupGlooAsyncTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main INSTALL_TEST ${INSTALL_TEST})
+    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main gtest INSTALL_TEST ${INSTALL_TEST})
+    c10d_add_test(ProcessGroupGlooAsyncTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main gtest INSTALL_TEST ${INSTALL_TEST})
+    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main gtest gloo INSTALL_TEST ${INSTALL_TEST})
+    c10d_add_test(ProcessGroupGlooAsyncTest.cpp LINK_LIBRARIES torch_cpu c10d_cuda_test gtest_main gtest gloo INSTALL_TEST ${INSTALL_TEST})
   endif()
   if(USE_NCCL AND USE_C10D_NCCL)
     # NCCL is a private dependency of libtorch, but the tests include some


@@ 451,7 463,7 @@ index 285a5dd2a73..51e76302265 100644
 else()
   if(USE_GLOO AND USE_C10D_GLOO)
-    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu gtest_main INSTALL_TEST OFF)
+    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu gtest_main gtest INSTALL_TEST OFF)
+    c10d_add_test(ProcessGroupGlooTest.cpp LINK_LIBRARIES torch_cpu gtest_main gtest gloo INSTALL_TEST OFF)
   endif()
 endif()
 


@@ 469,10 481,10 @@ index 14fd7f7ae9a..9564d1270cb 100644
   if(USE_CUDA)
     target_compile_definitions(test_dist_autograd PRIVATE "USE_CUDA")
diff --git i/test/cpp/jit/CMakeLists.txt w/test/cpp/jit/CMakeLists.txt
index d192d8a6c5d..2cdb54b9966 100644
index 0b2a06b53c9..9c04aec99a8 100644
--- i/test/cpp/jit/CMakeLists.txt
+++ w/test/cpp/jit/CMakeLists.txt
@@ -125,7 +125,7 @@ if(USE_MKLDNN)
@@ -126,7 +126,7 @@ if(USE_MKLDNN)
   target_link_libraries(test_jit PRIVATE caffe2::mkldnn)
 endif()
 


@@ 508,10 520,10 @@ index 5c27dd84f7c..27f7afdaeb4 100644
 if(LINUX)
   target_link_libraries(test_lite_interpreter_runtime PRIVATE "-Wl,--no-as-needed,$<TARGET_FILE:backend_with_compiler_runtime>,--as-needed")
diff --git i/test/cpp/nativert/CMakeLists.txt w/test/cpp/nativert/CMakeLists.txt
index 9f2ad858dfd..d264ed69216 100644
index 1b4752ed908..dc60673afbb 100644
--- i/test/cpp/nativert/CMakeLists.txt
+++ w/test/cpp/nativert/CMakeLists.txt
@@ -30,7 +30,7 @@ add_executable(test_nativert
@@ -58,7 +58,7 @@ add_executable(test_nativert
 # TODO temporary until we can delete the old gtest polyfills.
 target_compile_definitions(test_nativert PRIVATE USE_GTEST)
 


@@ 533,28 545,6 @@ index 454c54a0187..301319bbe71 100644
 )
 
 if(USE_GLOO)
diff --git i/test/cpp/tensorexpr/CMakeLists.txt w/test/cpp/tensorexpr/CMakeLists.txt
index 8fe6ffd525e..9650eee29e6 100644
--- i/test/cpp/tensorexpr/CMakeLists.txt
+++ w/test/cpp/tensorexpr/CMakeLists.txt
@@ -39,7 +39,7 @@ add_executable(test_tensorexpr
   ${TENSOREXPR_TEST_ROOT}/padded_buffer.cpp
   ${TENSOREXPR_TEST_SRCS})
 
-target_link_libraries(test_tensorexpr PRIVATE torch gtest_main)
+target_link_libraries(test_tensorexpr PRIVATE torch gtest_main gtest)
 target_include_directories(test_tensorexpr PRIVATE ${ATen_CPU_INCLUDE})
 target_compile_definitions(test_tensorexpr PRIVATE USE_GTEST)
 
@@ -51,7 +51,7 @@ target_include_directories(tutorial_tensorexpr PRIVATE ${ATen_CPU_INCLUDE})
 # pthreadpool header. For some build environment we need add the dependency
 # explicitly.
 if(USE_PTHREADPOOL)
-  target_link_libraries(test_tensorexpr PRIVATE pthreadpool_interface)
+  target_link_libraries(test_tensorexpr PRIVATE pthreadpool)
 endif()
 if(USE_CUDA)
   target_compile_definitions(test_tensorexpr PRIVATE USE_CUDA)
diff --git i/test/inductor/cpp/CMakeLists.txt w/test/inductor/cpp/CMakeLists.txt
index 78a1ad8c2fd..e350270bf40 100644
--- i/test/inductor/cpp/CMakeLists.txt


@@ 593,10 583,10 @@ index d38f0a41e22..136526b9dcd 100644
 target_compile_definitions(test_mobile_nnc PRIVATE USE_GTEST)
 
diff --git i/torch/CMakeLists.txt w/torch/CMakeLists.txt
index 3985cf50c14..51054202e95 100644
index 1632147f022..eff74feb501 100644
--- i/torch/CMakeLists.txt
+++ w/torch/CMakeLists.txt
@@ -83,9 +83,6 @@ set(TORCH_PYTHON_LINK_LIBRARIES
@@ -82,9 +82,6 @@ set(TORCH_PYTHON_LINK_LIBRARIES
     Python::Module
     pybind::pybind11
     opentelemetry::api