Skip to content

Conversation

@robinvandernoord
Copy link

I've updated the required dependencies (pyo3 and serde-pyobject) to versions that support Python 3.14.
This caused a compilation error when building, which I've fixed in src/python_bindings/mod.rs by adding a type hint.
Furthermore, I've added 3.14 as a build target in the build_wheels.yml github workflow.

Running make test still failed on 4 tests:

FAILED tests/test_kernels.py::test_interface_torch - RuntimeError: torch.compile is not supported on Python 3.14+
FAILED tests/test_kernels.py::test_interface_numpy - ImportError: To use the kernels in `outlines_core.kernels.numpy`, `numba` must be installed. You can install it with `pip install numba`
FAILED tests/test_kernels.py::test_torch_correctness - RuntimeError: torch.compile is not supported on Python 3.14+
FAILED tests/test_kernels.py::test_numpy_correctness - ImportError: To use the kernels in `outlines_core.kernels.numpy`, `numba` must be installed. You can install it with `pip install numba`

This is due to numba not supporting 3.14 and torch only partially supporting it.

@robinvandernoord
Copy link
Author

Installing latest numba via (uv) pip install git+https://github.com/numba/numba.git reduces the amount of errors to 2:

FAILED tests/test_kernels.py::test_interface_torch - RuntimeError: torch.compile is not supported on Python 3.14+
FAILED tests/test_kernels.py::test_torch_correctness - RuntimeError: torch.compile is not supported on Python 3.14+

This error is still raised on the latest main branch of torch: https://github.com/pytorch/pytorch/blob/main/torch/_dynamo/eval_frame.py#L1306

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant