-
Notifications
You must be signed in to change notification settings - Fork 61
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update numpy
version specifier
#480
Conversation
This PR should unblock rapidsai/dask-cuda#1080. |
I'm not sure if this is the right fix here. Only packages using Instead, shouldn't we be pinning only cudf (and not cucim)? https://github.com/rapidsai/cudf/blob/36b07f13480d8f7ba18274cb3a63f06855d848ff/conda/recipes/cudf/meta.yaml#L65 |
conda/recipes/cucim/meta.yaml
Outdated
@@ -43,7 +43,7 @@ requirements: | |||
- cudatoolkit ={{ cuda_version }} | |||
- cupy >=10,<12.0.0a0 | |||
- libcucim ={{ version }} | |||
- numpy >=1.21.3 | |||
- numpy >=1.21.3,<1.24 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Discussed with @ajschmidt8 offline. This was solving to 1.24.1
in the host
, which meant that the runtime pin_compatible
was something like >=1.24
-- which is incompatible with cudf.
The solution is that host
requirements should build against the oldest supported NumPy. Then the pin_compatible
won't conflict with cudf's requirements.
- numpy >=1.21.3,<1.24 | |
- numpy =1.21.3 |
Codecov ReportBase: 92.95% // Head: 92.95% // No change to project coverage 👍
Additional details and impacted files@@ Coverage Diff @@
## branch-23.02 #480 +/- ##
=============================================
Coverage 92.95% 92.95%
=============================================
Files 130 130
Lines 9775 9775
=============================================
Hits 9086 9086
Misses 689 689 Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
Co-authored-by: jakirkham <[email protected]>
/merge |
Thanks all! 🙏 |
This PR updates the
numpy
version specifier forcucim
.It was previously updated in this commit to enable publishing Python
3.10
packages.However, since there was no maximum version in that specifier,
conda
was picking up the latest version ofnumpy
(1.24.1
) which is incompatible withcudf
's version ofnumba
(numba>=0.56.2
).For instance, if you try to run this:
mamba create -n test 'numba>=0.56.2' 'numpy>=1.24.1'
, it will fail to solve.Running with the new version specifier in this PR does solve:
mamba create -n test 'numba>=0.56.2' 'numpy>=1.21.3,<1.24'
.Therefore, this maximum version limit seems necessary for
cucim
to be compatible with the rest of RAPIDS.