Drop 3.10 #2
Drop 3.10 #2
Conversation
There was a problem hiding this comment.
Pull Request Overview
This PR drops support for Python 3.10 and older NumPy versions (< 2.0), standardizing the codebase on NumPy 2.x only. The changes remove compatibility code that was needed to support both NumPy 1.x and 2.x versions.
- Removes the
pytensor.npy_2_compatmodule which provided compatibility abstractions - Updates imports to use NumPy 2.x native functions directly
- Updates minimum version requirements across all configuration files
Reviewed Changes
Copilot reviewed 28 out of 28 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
pytensor/npy_2_compat.py |
Removes most compatibility code, keeping only the old_np_unique function |
| Multiple Python files | Replace compatibility imports with direct NumPy 2.x imports |
| Test files | Remove conditional NumPy version checks and update exception handling |
| Configuration files | Update Python and NumPy version requirements to 3.11+ and 2.0+ |
| CI workflows | Update test matrices to remove Python 3.10 and NumPy version variations |
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
| The value of the numpy C API NPY_RAVEL_AXIS. | ||
| """ | ||
|
|
||
| NDARRAY_C_VERSION = np._core._multiarray_umath._get_ndarray_c_version() # type: ignore[attr-defined] |
There was a problem hiding this comment.
Direct access to private NumPy attributes (np._core._multiarray_umath._get_ndarray_c_version()) is fragile and may break in future NumPy releases. Consider using a more stable public API or adding error handling for when this private attribute is unavailable.
| NDARRAY_C_VERSION = np._core._multiarray_umath._get_ndarray_c_version() # type: ignore[attr-defined] | |
| try: | |
| NDARRAY_C_VERSION = np._core._multiarray_umath._get_ndarray_c_version() # type: ignore[attr-defined] | |
| except AttributeError: | |
| logging.warning("Could not access np._core._multiarray_umath._get_ndarray_c_version(); NDARRAY_C_VERSION set to None.") | |
| NDARRAY_C_VERSION = None |
Copilot uses AI. Check for mistakes.
| ) | ||
|
|
||
| def test_too_big_rank(self): | ||
| numpy_maxdims = 64 |
There was a problem hiding this comment.
The hardcoded value 64 should be replaced with a named constant from a central location (like pytensor.utils) to avoid magic numbers and ensure consistency across the codebase.
| numpy_maxdims = 64 | |
| numpy_maxdims = np.MAXDIMS |
Copilot uses AI. Check for mistakes.
| from numpy._core.einsumfunc import ( # type: ignore[attr-defined] | ||
| _find_contraction, | ||
| _parse_einsum_input, | ||
| normalize_axis_index, | ||
| normalize_axis_tuple, | ||
| ) |
There was a problem hiding this comment.
Direct import from private NumPy modules (numpy._core.einsumfunc) is fragile and may break in future NumPy releases. Consider using public APIs or adding fallback imports with error handling.
Copilot uses AI. Check for mistakes.
2 participants
📚 Documentation preview 📚: https://pytensor--2.org.readthedocs.build/en/2/