Nightly python packages for python 3.14#26397
Conversation
tools/ci_build/github/azure-pipelines/templates/py-package-smoking-test.yml
Outdated
Show resolved
Hide resolved
521677e to
010bc28
Compare
If you mean the A10 machines, they use a special driver from https://learn.microsoft.com/en-us/azure/virtual-machines/windows/n-series-driver-setup Now the driver is preinstalled in the VM images. But, now our infra has the ability to help us auto install the driver when VM starts up. We can use that feature then we won't need to remember to update the drivers. So, I need to create a new VM image without driver and test it. |
Right. Updating driver is independent of python 3.14. I think we can update the driver in another PR. |
This is a follow up change for #26397: (1) Enable python 3.14 for webgpu package (2) Skip python tests for python 3.14 just like in python cuda package pipelines. onnx package does not support python 3.14 right now, so python tests depending on onnx will fail. Example error in python package pipeline: ``` Traceback (most recent call last): File "E:\_work\1\s\build\Release\Release\onnx_backend_test_series.py", line 14, in <module> import onnx ModuleNotFoundError: No module named 'onnx' ``` This is a temporary walkaround to unblock python packaging pipeline for python 3.14. We can re-enable the test after onnx next release (in a few weeks?)
Missed it in #26397. Add it to make the cpu python package complete.
Update python packaging pipeline to support python 3.14. This is to add python 3.14 wheels to both CUDA 12 and CUDA 13 nightly packaging pipelines. The CUDA 13 pipelines were added in microsoft#26420. For linux, two wheels are added for python 3.14 and 3.14t. For Windows, a wheel is added for pyhthon 3.14. Known issues: 1. onnx does not have prebuild package for python 3.14, which causes some pipelines failed. 2. onnxscript and onnx-ir depends on onnx. 3. The CI machine for Windows cuda python packaging has cuda driver 553.24, which does not support CUDA 13.0. TODO: Update cuda driver to 580 or later to test properly. Walkarounds for issues 1 and 2: * For python 3.14, we skip onnx, onnxscript and onnx-ir in requirements.txt. * Disable related tests. When onnx package supports python 3.14 in the future, we can undo the walkarounds and re-enable those tests.
This is a follow up change for microsoft#26397: (1) Enable python 3.14 for webgpu package (2) Skip python tests for python 3.14 just like in python cuda package pipelines. onnx package does not support python 3.14 right now, so python tests depending on onnx will fail. Example error in python package pipeline: ``` Traceback (most recent call last): File "E:\_work\1\s\build\Release\Release\onnx_backend_test_series.py", line 14, in <module> import onnx ModuleNotFoundError: No module named 'onnx' ``` This is a temporary walkaround to unblock python packaging pipeline for python 3.14. We can re-enable the test after onnx next release (in a few weeks?)
Missed it in microsoft#26397. Add it to make the cpu python package complete.
### Description The requires-python field in the [metadata](https://pypi.org/pypi/onnxruntime/json) is currently set to `python >= 3.10`. However, since ONNX Runtime has dropped support for `Python 3.10`, this constraint is now inaccurate and may lead to dependency resolution issues in `uv lock`. Related Commits: #26397
### Description The requires-python field in the [metadata](https://pypi.org/pypi/onnxruntime/json) is currently set to `python >= 3.10`. However, since ONNX Runtime has dropped support for `Python 3.10`, this constraint is now inaccurate and may lead to dependency resolution issues in `uv lock`. Related Commits: #26397
Update python packaging pipeline to support python 3.14. This is to add python 3.14 wheels to both CUDA 12 and CUDA 13 nightly packaging pipelines. The CUDA 13 pipelines were added in #26420.
For linux, two wheels are added for python 3.14 and 3.14t.
For Windows, a wheel is added for pyhthon 3.14.
Known issues:
Walkarounds for issues 1 and 2:
When onnx package supports python 3.14 in the future, we can undo the walkarounds and re-enable those tests.