-
Notifications
You must be signed in to change notification settings - Fork 59
Closed
Labels
bugSomething isn't workingSomething isn't working
Description
Describe the bug
The same tests pass with NumPy 2.0.1 but fail with 2.2.1.
Upon a closer inspection, I noticed that somewhere in this repo NumPy has been pinned to 2.0.2, so all CI runs in this repo could not catch this test error. However, I am unable to locate where the pinning is done. It could be possible that it comes from the RAPIDS CI setup, since lots of CI infra here seem to come from RAPIDS.
Steps/Code to reproduce bug
Update NumPy to 2.2.x and rerun the tests.
Expected behavior
Test pass.
Environment details (please complete the following information):
See below.
Additional context
This is a spin-off of nvbug 5011593. Below is what QA reported:
(numba_py310) C:\tmp\cuda-python-public\cuda_bindings>pip install numpy -U
Requirement already satisfied: numpy in c:\programdata\miniconda3\envs\numba_py310\lib\site-packages (2.0.1)
Collecting numpy
Using cached numpy-2.2.1-cp310-cp310-win_amd64.whl.metadata (60 kB)
Using cached numpy-2.2.1-cp310-cp310-win_amd64.whl (12.9 MB)
Installing collected packages: numpy
Attempting uninstall: numpy
Found existing installation: numpy 2.0.1
Uninstalling numpy-2.0.1:
Successfully uninstalled numpy-2.0.1
Successfully installed numpy-2.2.1
(numba_py310) C:\tmp\cuda-python-public\cuda_bindings>set NUMBA_CUDA_USE_NVIDIA_BINDING=1 && python -m numba.runtests numba.cuda.tests
........s.........s........................s.s................................................................................................................................................................................................................................................................................................................................................................................................................ss.s...............................s......EE.....................................................................................................................s..s........................................sssssssssssssssssssss...s.........s..............s....................................................................................................................................C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\tests\cudapy\test_atomics.py:673: RuntimeWarning: overflow encountered in scalar subtract
gold[orig[i]] -= 1
................................s....................................................................ss.............................s................x............x..............................x.......x..........................................................s............................................................................sss.ss............................................................C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\tests\cudapy\test_math.py:620: RuntimeWarning: overflow encountered in cast
Cref[i] = math.pow(A[i], B[i])
C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\tests\cudapy\test_math.py:621: RuntimeWarning: overflow encountered in cast
np.testing.assert_allclose(np.power(A, B).astype(npdtype), C, rtol=1e-6)
.........................ss...C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\dispatcher.py:597: NumbaPerformanceWarning: Grid size 1 will likely result in GPU under-utilization due to low occupancy.
warn(NumbaPerformanceWarning(msg))
....................................................................x...xxxxx..................................sx.s......................................................s.................................................................s......................s....ss........
======================================================================
ERROR: test_empty_slice_1d (numba.cuda.tests.cudadrv.test_cuda_array_slicing.CudaArraySlicing)
----------------------------------------------------------------------
Traceback (most recent call last):
File "C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\tests\cudadrv\test_cuda_array_slicing.py", line 189, in test_empty_slice_1d
self.assertFalse(darr[:0][:0].copy_to_host())
File "C:\ProgramData\miniconda3\envs\numba_py310\lib\unittest\case.py", line 679, in assertFalse
if expr:
ValueError: The truth value of an empty array is ambiguous. Use `array.size > 0` to check that an array is not empty.
======================================================================
ERROR: test_empty_slice_2d (numba.cuda.tests.cudadrv.test_cuda_array_slicing.CudaArraySlicing)
----------------------------------------------------------------------
Traceback (most recent call last):
File "C:\ProgramData\miniconda3\envs\numba_py310\lib\site-packages\numba_cuda\numba\cuda\tests\cudadrv\test_cuda_array_slicing.py", line 202, in test_empty_slice_2d
self.assertFalse(darr[:0][:0].copy_to_host())
File "C:\ProgramData\miniconda3\envs\numba_py310\lib\unittest\case.py", line 679, in assertFalse
if expr:
ValueError: The truth value of an empty array is ambiguous. Use `array.size > 0` to check that an array is not empty.
----------------------------------------------------------------------
Ran 1539 tests in 468.728s
FAILED (errors=2, skipped=53, expected failures=11)
(numba_py310) C:\tmp\cuda-python-public\cuda_bindings>pip list
Package Version Editable project location
---------------- ------------------------------ ---------------------------------------
colorama 0.4.6
cuda-python 12.6.2.post1+16.ge09b7fb c:\tmp\cuda-python-public\cuda_bindings
Cython 3.0.11
exceptiongroup 1.2.2
iniconfig 2.0.0
llvmlite 0.45.0.dev0+0.ge9f12c6.dirty
mkl_fft 1.3.11
mkl_random 1.2.8
mkl-service 2.4.0
numba 0.62.0.dev0+8.g956ea6c69.dirty
numba-cuda 0.3.0
numpy 2.2.1
packaging 24.2
pip 24.2
pluggy 1.5.0
py-cpuinfo 9.0.0
pyclibrary 0.2.2
pyparsing 3.2.1
pytest 8.3.4
pytest-benchmark 5.1.0
pywin32 308
setuptools 75.1.0
tomli 2.2.1
versioneer 0.29
wheel 0.44.0
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working