ci: refactor cpu wheel build workflow by Bing-su · Pull Request #2164 · abetlen/llama-cpp-python

Bing-su · 2026-03-26T13:09:06Z

Builds widely reusable wheels.

In fact, since llama-cpp-python uses ctypes, there is no need to specify a Python ABI. Therefore, tags like py3-none-manylinux_2_34_x86_64.whl can be used.

Please check the build success log at this link:
https://github.com/Bing-su/llama-cpp-python/actions/runs/23591453286

….toml

Bing-su · 2026-03-26T13:11:01Z

.github/workflows/build-and-release.yaml

    strategy:
      matrix:
-        os: [ubuntu-22.04, windows-2022, macos-14, macos-15]
+        os: [ubuntu-22.04, windows-2022, macos-15]


macos-15-intel failed to build, so it was excluded.
https://github.com/Bing-su/llama-cpp-python/actions/runs/23578504562/job/68656271494

Bing-su · 2026-03-26T13:12:31Z

pyproject.toml

@@ -1,5 +1,5 @@
 [build-system]
-requires = ["scikit-build-core[pyproject]>=0.9.2"]
+requires = ["scikit-build-core[wheels]>=0.12.2"]


scikit-build-core[pyproject] is now a backward-compatibility-only extra.

[project.optional-dependencies] pyproject = [ ] wheels = [ "cmake", "ninja; sys_platform!='win32'", ] wheel-free-setuptools = [ 'setuptools>=70.1; python_version>="3.8"', ]

Bing-su · 2026-03-26T13:14:40Z

pyproject.toml


 [tool.scikit-build]
 wheel.packages = ["llama_cpp"]
+wheel.py-api = "py3"


This option generates tags like py3-none-manylinux_2_34_x86_64.whl.
https://scikit-build-core.readthedocs.io/en/latest/reference/configs.html#confval-wheel.py-api

Bing-su · 2026-03-26T13:15:52Z

pyproject.toml

+# see all manylinux images here: https://github.com/pypa/manylinux
+manylinux-x86_64-image = "manylinux_2_34"
+manylinux-aarch64-image = "manylinux_2_34"
+skip = "*musllinux*"  # Segmentation fault occurs on musllinux python 3.10+


Successfully installed MarkupSafe-3.0.3 annotated-doc-0.0.4 annotated-types-0.7.0 anyio-4.13.0 certifi-2026.2.25 click-8.3.1 diskcache-5.6.3 exceptiongroup-1.3.1 fastapi-0.135.2 filelock-3.25.2 fsspec-2026.2.0 h11-0.16.0 hf-xet-1.4.2 httpcore-1.0.9 httpx-0.28.1 huggingface-hub-1.8.0 idna-3.11 iniconfig-2.3.0 jinja2-3.1.6 llama-cpp-python-0.3.19 markdown-it-py-4.0.0 mdurl-0.1.2 numpy-2.2.6 packaging-26.0 pluggy-1.6.0 pydantic-2.12.5 pydantic-core-2.41.5 pydantic-settings-2.13.1 pygments-2.19.2 pytest-9.0.2 python-dotenv-1.2.2 pyyaml-6.0.3 rich-14.3.3 scipy-1.15.3 shellingham-1.5.4 sse-starlette-3.3.3 starlette-1.0.0 starlette-context-0.3.6 tomli-2.4.1 tqdm-4.67.3 typer-0.24.1 typing-extensions-4.15.0 typing-inspection-0.4.2 + mkdir -p /tmp/tmp.ACiMCD/test_cwd + mkdir -p /tmp/tmp.ACiMCD/test_cwd/tests + mkdir -p /tmp/tmp.ACiMCD/test_cwd/vendor + sh -c 'pytest -v ./tests' ============================= test session starts ============================== platform linux -- Python 3.10.19, pytest-9.0.2, pluggy-1.6.0 -- /tmp/tmp.ACiMCD/venv/bin/python cachedir: .pytest_cache rootdir: /tmp/tmp.ACiMCD/test_cwd plugins: anyio-4.13.0 collecting ... collected 11 items tests/test_llama.py::test_llama_cpp_version PASSED [ 9%] tests/test_llama.py::test_llama_cpp_tokenization PASSED [ 18%] tests/test_llama.py::test_real_model PASSED [ 27%] Fatal Python error: Segmentation fault Thread 0x00007fe9f04f6b30 (most recent call first): File "/opt/python/cp310-cp310/lib/python3.10/threading.py", line 324 in wait File "/opt/python/cp310-cp310/lib/python3.10/threading.py", line 607 in wait File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/tqdm/_monitor.py", line 60 in run File "/opt/python/cp310-cp310/lib/python3.10/threading.py", line 1016 in _bootstrap_inner File "/opt/python/cp310-cp310/lib/python3.10/threading.py", line 973 in _bootstrap Current thread 0x00007fea0c0bab28 (most recent call first): Garbage-collecting File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/llama_cpp/_internals.py", line 273 in free_ctx File "/opt/python/cp310-cp310/lib/python3.10/contextlib.py", line 449 in _exit_wrapper File "/opt/python/cp310-cp310/lib/python3.10/contextlib.py", line 561 in __exit__ File "/opt/python/cp310-cp310/lib/python3.10/contextlib.py", line 584 in close File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/llama_cpp/_internals.py", line 279 in close File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/llama_cpp/_internals.py", line 282 in __del__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 721 in parse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 731 in parse_tuple File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1021 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 214 in parse_statements File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 260 in parse_if File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 177 in parse_statement File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1030 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 214 in parse_statements File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 260 in parse_if File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 177 in parse_statement File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1030 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 214 in parse_statements File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 248 in parse_for File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 177 in parse_statement File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1030 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 214 in parse_statements File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 260 in parse_if File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 177 in parse_statement File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1030 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 214 in parse_statements File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 447 in parse_macro File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 177 in parse_statement File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1030 in subparse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/parser.py", line 1047 in parse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/environment.py", line 622 in _parse File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/environment.py", line 763 in compile File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/jinja2/environment.py", line 1111 in from_string File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/llama_cpp/llama_chat_format.py", line 216 in __init__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/llama_cpp/llama.py", line 516 in __init__ File "/tmp/tmp.ACiMCD/test_cwd/tests/test_llama.py", line 135 in test_real_llama File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/python.py", line 166 in pytest_pyfunc_call File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_callers.py", line 121 in _multicall File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_manager.py", line 120 in _hookexec File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_hooks.py", line 512 in __call__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/python.py", line 1720 in runtest File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 179 in pytest_runtest_call File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_callers.py", line 121 in _multicall File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_manager.py", line 120 in _hookexec File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_hooks.py", line 512 in __call__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 245 in <lambda> File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 353 in from_call File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 244 in call_and_report File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 137 in runtestprotocol File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/runner.py", line 118 in pytest_runtest_protocol File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_callers.py", line 121 in _multicall File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_manager.py", line 120 in _hookexec File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_hooks.py", line 512 in __call__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/main.py", line 396 in pytest_runtestloop File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_callers.py", line 121 in _multicall File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_manager.py", line 120 in _hookexec File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_hooks.py", line 512 in __call__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/main.py", line 372 in _main File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/main.py", line 318 in wrap_session File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/main.py", line 365 in pytest_cmdline_main File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_callers.py", line 121 in _multicall File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_manager.py", line 120 in _hookexec File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/pluggy/_hooks.py", line 512 in __call__ File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/config/__init__.py", line 199 in main File "/tmp/tmp.ACiMCD/venv/lib/python3.10/site-packages/_pytest/config/__init__.py", line 223 in console_main File "/tmp/tmp.ACiMCD/venv/bin/pytest", line 6 in <module> Extension modules: numpy._core._multiarray_umath, numpy.linalg._umath_linalg, scipy._lib._ccallback_c, scipy.special._ufuncs_cxx, scipy.special._ufuncs, numpy.random._common, numpy.random.bit_generator, numpy.random._bounded_integers, numpy.random._mt19937, numpy.random.mtrand, numpy.random._philox, numpy.random._pcg64, numpy.random._sfc64, numpy.random._generator, scipy.special._specfun, scipy.special._comb, scipy.linalg._fblas, scipy.linalg._flapack, scipy.linalg.cython_lapack, scipy.linalg._cythonized_array_utils, scipy.linalg._solve_toeplitz, scipy.linalg._decomp_lu_cython, scipy.linalg._matfuncs_sqrtm_triu, scipy.linalg._matfuncs_expm, scipy.linalg._linalg_pythran, scipy.linalg.cython_blas, scipy.linalg._decomp_update, scipy.sparse._sparsetools, _csparsetools, scipy.sparse._csparsetools, scipy.sparse.linalg._dsolve._superlu, scipy.sparse.linalg._eigen.arpack._arpack, scipy.sparse.linalg._propack._spropack, scipy.sparse.linalg._propack._dpropack, scipy.sparse.linalg._propack._cpropack, scipy.sparse.linalg._propack._zpropack, scipy.sparse.csgraph._tools, scipy.sparse.csgraph._shortest_path, scipy.sparse.csgraph._traversal, scipy.sparse.csgraph._min_spanning_tree, scipy.sparse.csgraph._flow, scipy.sparse.csgraph._matching, scipy.sparse.csgraph._reordering, scipy.special._ellip_harm_2, yaml._yaml, markupsafe._speedups (total: 46) /bin/bash: line 258: 1252 Segmentation fault (core dumped) env CIBUILDWHEEL=1 PKG_CONFIG_PATH=/usr/local/lib/pkgconfig HOSTNAME=168e6e417a38 LANGUAGE=en_US.UTF-8 AUDITWHEEL_POLICY=musllinux_1_2 AUDITWHEEL_ARCH=x86_64 *** HOME=/root LANG=en_US.UTF-8 SHLVL=1 LD_LIBRARY_PATH='' SSL_CERT_FILE=/opt/_internal/certs.pem AUDITWHEEL_PLAT=musllinux_1_2_x86_64 LC_ALL=en_US.UTF-8 PATH=/tmp/tmp.ACiMCD/venv/bin:/opt/python/cp310-cp310/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin DEVTOOLSET_ROOTPATH='' OLDPWD=/ _=/usr/bin/env PIP_DISABLE_PIP_VERSION_CHECK=1 PIP_ROOT_USER_ACTION=ignore CMAKE_ARGS='-DGGML_NATIVE=off -DBUILD_SHARED_LIBS=OFF' VIRTUAL_ENV=/tmp/tmp.ACiMCD/venv sh -c 'pytest -v ./tests' tests/test_llama.py::test_real_llama ::endgroup::Testing wheel...

Bing-su · 2026-03-26T13:20:06Z

pyproject.toml

+
+[tool.cibuildwheel]
+archs = ["auto64"]
+# llama.cpp requires glibc 2.31+. manylinux_2_34 is the oldest manylinux image with glibc 2.31.


INFO:auditwheel.wheel_abi:setting libc to glibc llama_cpp_python-0.3.19-py3-none-linux_x86_64.whl is consistent with the following platform tag: "linux_x86_64". The wheel references external versioned symbols in these system-provided shared libraries: libgcc_s.so.1 with versions {'GCC_3.0', 'GCC_3.4'}, libm.so.6 with versions {'GLIBC_2.27', 'GLIBC_2.2.5'}, libstdc++.so.6 with versions {'GLIBCXX_3.4.17', 'CXXABI_1.3.8', 'CXXABI_1.3.5', 'GLIBCXX_3.4.19', 'CXXABI_1.3', 'GLIBCXX_3.4.18', 'GLIBCXX_3.4.25', 'CXXABI_1.3.11', 'GLIBCXX_3.4.22', 'GLIBCXX_3.4.9', 'GLIBCXX_3.4.14', 'GLIBCXX_3.4.21', 'GLIBCXX_3.4.15', 'CXXABI_1.3.2', 'CXXABI_1.3.9', 'GLIBCXX_3.4', 'GLIBCXX_3.4.20', 'GLIBCXX_3.4.11', 'CXXABI_1.3.3'}, libc.so.6 with versions {'GLIBC_2.17', 'GLIBC_2.16', 'GLIBC_2.14', 'GLIBC_2.3', 'GLIBC_2.6', 'GLIBC_2.4', 'GLIBC_2.7', 'GLIBC_2.2.5'}, libpthread.so.0 with versions {'GLIBC_2.3.4', 'GLIBC_2.2.5'}, libgomp.so.1 with versions {'GOMP_1.0', 'GOMP_4.0', 'OMP_1.0'}, libdl.so.2 with versions {'GLIBC_2.2.5'} This constrains the platform tag to "manylinux_2_31_x86_64". In order to achieve a more compatible tag, you would need to recompile a new wheel from source on a system with earlier versions of these libraries, such as a recent manylinux image. INFO:auditwheel.main_repair:Repairing llama_cpp_python-0.3.19-py3-none-linux_x86_64.whl usage: auditwheel [-h] [-V] [-v] command ... auditwheel: error: cannot repair "/tmp/cibuildwheel/built_wheel/llama_cpp_python-0.3.19-py3-none-linux_x86_64.whl" to "manylinux_2_28_x86_64" ABI because of the presence of too-recent versioned symbols. You'll need to compile the wheel on an older toolchain.

The [wheels] extra is unnecessary

abetlen · 2026-03-29T05:48:10Z

@Bing-su thank you for the contribution, these are some really good changes.

I'm going to likely break this up across a few different PRs to ensure it's easy to rollback in case anything goes wrong but I'll be sure to credit you in the CHANGELOG.

ci: refactor wheel build workflow and move configuration to pyproject…

e2cb5a4

….toml

Bing-su commented Mar 26, 2026

View reviewed changes

Bing-su added 2 commits March 26, 2026 22:32

ci: restore Install dependencies step

0c8fe87

ci: remove wheels extra

4309743

The [wheels] extra is unnecessary

abetlen mentioned this pull request Mar 29, 2026

ci: publish release wheels as py3-none #2166

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ci: refactor cpu wheel build workflow#2164

ci: refactor cpu wheel build workflow#2164
Bing-su wants to merge 3 commits intoabetlen:mainfrom
Bing-su:ci/cpu

Bing-su commented Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026 •

edited

Loading

Uh oh!

abetlen commented Mar 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

Bing-su commented Mar 26, 2026

Uh oh!

Bing-su Mar 26, 2026

Choose a reason for hiding this comment

Uh oh!

Bing-su Mar 26, 2026

Choose a reason for hiding this comment

Uh oh!

Bing-su Mar 26, 2026

Choose a reason for hiding this comment

Uh oh!

Bing-su Mar 26, 2026

Choose a reason for hiding this comment

Uh oh!

Bing-su Mar 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

abetlen commented Mar 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Bing-su Mar 26, 2026 •

edited

Loading