Drop {{ pin_compatible('numpy', max_pin='x') }}
#16301
Conversation
jakirkham
added
improvement
jakirkham
force-pushed
the
drop_np_pin_compat
branch
4 times, most recently
from
58cf117
to
774f66b
Compare
This is no longer needed as `numpy`'s own `run_exports` sets this for us. In `requirements/run` set the `numpy` lower bound to `1.23` so that we have a higher bound than the `run_exports` would otherwise use.
jakirkham
force-pushed
the
drop_np_pin_compat
branch
from
774f66b
to
74781b0
Compare
| # TODO: Change to `2.0` for NumPy 2 | ||
| - numpy 1.23 |
There was a problem hiding this comment.
Once we update to NumPy 2, that will be true
Right now we build with NumPy 1.23, which means run_exports adds {{ pin_subpackage("numpy") }}. The defaults of pin_subpackage are to tightly pin on the lower bound and pin the upper bound based on major version. So this turns into numpy >=1.23.5,<2.0a0
There was a problem hiding this comment.
oh got it, right. Ok sounds good to me.
| # TODO: Update `numpy` in `host` when dropping `<2.0a0` | ||
| - numpy >=1.23,<2.0a0 |
There was a problem hiding this comment.
While this seems duplicative to what numpy's run_exports does, this same pinning is used throughout RAPIDS. Also it gives a foothold for RDFG changes to be made to move to NumPy 2. When that does happen, we have a comment above to remind us to make other needed changes in this recipe
|
/merge |
|
Thanks James! 🙏 |
Description
Part of issue: rapidsai/build-planning#82
Drop
{{ pin_compatible('numpy', max_pin='x') }}as it is no longer needed.numpyhas its ownrun_exports, which constraintsnumpyto an API compatible version. More details in issue: https://github.com/orgs/rapidsai/projects/132So
cudfnow uses that in its recipe builds. Also updaterequirements/runto set thenumpylower bound to1.23as required by us.Lastly add todo comments for NumPy 2 update lines.
Checklist