[Code scan] Python builds should discover Torch CMake prefix when BUILD_PY_IF is true

[njzjz]

This issue was found by a Codex global scan of the repository at commit 19f9265.

pyproject.toml configures scikit-build to build the Python interface only:

deepmd-gnn/pyproject.toml

Lines 97 to 99 in 19f9265

	[tool.scikit-build.cmake.define]
	BUILD_PY_IF = true
	BUILD_CPP_IF = false

However, the CMake logic that queries torch.utils.cmake_prefix_path only runs for selected C++ builds or CIBUILDWHEEL=1:

deepmd-gnn/CMakeLists.txt

Lines 85 to 109 in 19f9265

	if(BUILD_CPP_IF
	AND USE_PT_PYTHON_LIBS
	AND NOT CMAKE_CROSSCOMPILING
	AND NOT SKBUILD
	OR "$ENV{CIBUILDWHEEL}" STREQUAL "1")
	find_package(
	Python
	COMPONENTS Interpreter
	REQUIRED)
	execute_process(
	COMMAND ${Python_EXECUTABLE} -c
	"import torch;print(torch.utils.cmake_prefix_path)"
	WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
	OUTPUT_VARIABLE PYTORCH_CMAKE_PREFIX_PATH
	RESULT_VARIABLE PYTORCH_CMAKE_PREFIX_PATH_RESULT_VAR
	ERROR_VARIABLE PYTORCH_CMAKE_PREFIX_PATH_ERROR_VAR
	OUTPUT_STRIP_TRAILING_WHITESPACE)
	if(NOT ${PYTORCH_CMAKE_PREFIX_PATH_RESULT_VAR} EQUAL 0)
	message(
	FATAL_ERROR
	"Cannot determine PyTorch CMake prefix path, error code: $PYTORCH_CMAKE_PREFIX_PATH_RESULT_VAR}, error message: ${PYTORCH_CMAKE_PREFIX_PATH_ERROR_VAR}"
	)
	endif()
	list(APPEND CMAKE_PREFIX_PATH ${PYTORCH_CMAKE_PREFIX_PATH})
	endif()

find_package(Torch REQUIRED) then runs unconditionally:

deepmd-gnn/CMakeLists.txt

Lines 120 to 124 in 19f9265

	find_package(Torch REQUIRED)
	if(Torch_VERSION VERSION_LESS "2.10.0")
	message(FATAL_ERROR "deepmd-gnn OP requires PyTorch >= 2.10.0 for the "
	"LibTorch Stable ABI.")
	endif()

A normal local configure without user-supplied CMAKE_PREFIX_PATH fails with:

Could not find a package configuration file provided by "Torch"

Suggested fix: when BUILD_PY_IF is true and Torch_DIR/CMAKE_PREFIX_PATH is not already sufficient, query torch.utils.cmake_prefix_path using the build Python interpreter before find_package(Torch REQUIRED).