Build and test with CUDA 13.0.0

[jameslamb]

Contributes to rapidsai/build-planning#208

• uses CUDA 13.0.0 to build and test
• adds CUDA 13 devcontainers
• moves some dependency pins (most others were done in use 'nvidia-ml-py' instead of 'pynvml', declare 'numba-cuda' dependency pins #7164)
  • cuda-python: >=13.0.1 (CUDA 13)

Contributes to rapidsai/build-planning#68

• updates to CUDA 13 dependencies in fallback entries in dependencies.yaml matrices (i.e., the ones that get written to pyproject.toml in source control)

Notes for Reviewers

This switches GitHub Actions workflows to the cuda13.0 branch from here: rapidsai/shared-workflows#413

A future round of PRs will revert that back to branch-25.10, once all of RAPIDS supports CUDA 13.

[robertmaynard]

#6823 was merged yesterday and added a dependency to pynvml constrainted to the 12.X series. We will need to update this PR with latest 25.10 and update that logic for CTK 12 and 13 @jameslamb

[jameslamb]

Problem 1: missing "treebank" dataset

update: seemed to be a network error, not seen on re-runs.

details (click me)

One CUDA 12.0.1 conda-python-tests-singlegpu job failed like this:

[gw1] linux -- Python 3.12.11 /opt/conda/envs/test/bin/python
Traceback (most recent call last):
  File "/opt/conda/envs/test/lib/python3.12/site-packages/nltk/corpus/util.py", line 84, in __load
    root = nltk.data.find(f"{self.subdir}/{zip_name}")
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/envs/test/lib/python3.12/site-packages/nltk/data.py", line 579, in find
    raise LookupError(resource_not_found)
LookupError: 
**********************************************************************
  Resource treebank not found.
  Please use the NLTK Downloader to obtain the resource:

  >>> import nltk

(conda-python-tests-singlegpu build link)

Hopefully just a temporary issue that's be resolved by a re-run.

Problem 2: Failing Naive Bayes tests (conda)

update: Related to scikit-learn dataset downloads interacting in unexpected way with tests run in parallel. Fixed by #7169

details (click me)

CUDA 13 conda-python-tests jobs both failed like this:

FAILED test_naive_bayes.py::test_gaussian_parameters[1e-05-balanced] - ValueError: Number of priors must match number of classes.
FAILED test_naive_bayes.py::test_gaussian_parameters[1e-05-unbalanced] - ValueError: Number of priors must match number of classes.
..
FAILED test_naive_bayes.py::test_categorical_partial_fit[True-int32-float32] - assert 0.1452 <= (0.104 + 0.0001)
FAILED test_naive_bayes.py::test_categorical_partial_fit[True-int32-float64] - assert 0.1452 <= (0.104 + 0.0001)
..
FAILED test_naive_bayes.py::test_categorical_parameters[False-False-0.1-balanced] - ValueError: Number of classes must match number of priors
FAILED test_naive_bayes.py::test_categorical_parameters[False-False-0.1-unbalanced] - ValueError: Number of classes must match number of priors
..
= 40 failed, 14374 passed, 6142 skipped, 1225 xfailed, 24 xpassed, 1225 warnings in 4949.54s (1:22:29) =

(conda-python-tests build link)

tracked in #7152

Problem 3: Failing Naive Bayes and Logistic Regression tests (wheels)

update: same as above, fixed by #7169

details (click me)

All of the amd64 CUDA 13 wheels tests passed, but on arm64 there were some test failures.

FAILED test_naive_bayes.py::test_multinomial_partial_fit[int32-float32] - assert 0.9238375200427579 >= 0.924
 +  where 0.9238375200427579 = accuracy_score(array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,), dtype=int32), array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,), dtype=int32))
FAILED test_naive_bayes.py::test_multinomial_partial_fit[int32-float64] - assert 0.9238375200427579 >= 0.924
 +  where 0.9238375200427579 = accuracy_score(array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,), dtype=int32), array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,)))
...
FAILED test_naive_bayes.py::test_multinomial[int32-float32] - assert 0.9238375200427579 >= 0.924
 +  where 0.9238375200427579 = accuracy_score(array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,)), array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,), dtype=int32))
FAILED test_naive_bayes.py::test_multinomial[int32-float64] - assert 0.9238375200427579 >= 0.924
 +  where 0.9238375200427579 = accuracy_score(array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,)), array([ 5, 16,  6, ..., 14,  7,  3], shape=(11226,)))
...
FAILED test_naive_bayes.py::test_gaussian_partial_fit - assert 0.988 >= 0.99
 +  where 0.988 = accuracy_score(array([ 5, 16,  6, ...,  8, 13, 11], shape=(1500,), dtype=int32), array([ 5, 16,  6, ...,  8, 13, 11], shape=(1500,), dtype=int32))
...
FAILED test_naive_bayes.py::test_categorical_partial_fit[False-int64-float32] - assert 0.1082 <= (0.104 + 0.0001)
FAILED test_naive_bayes.py::test_categorical_partial_fit[False-int64-float64] - assert 0.1082 <= (0.104 + 0.0001)
FAILED test_linear_model.py::test_logistic_regression_sparse_only - ExceptionGroup: Hypothesis found 2 distinct failures in explicit examples. (2 sub-exceptions)
= 22 failed, 14354 passed, 6150 skipped, 1223 xfailed, 24 xpassed, 1217 warnings in 1611.55s (0:26:51) =

(wheel-tests-cuml build link)

tracked in #7152 and #7162

[jameslamb]

/ok to test

[jameslamb]

/ok to test

[csadorf]

LGTM! Thanks a lot!

[jameslamb]

/merge