Unit test fails with NumPy 2.3.1 but passes with NumPy 1.26.4

[IamTingTing]

Describe the bug A unit test tests.transforms.test_signal_remove_frequency.TestSignalRemoveFrequencyNumpy.test_correct_parameters_multi_channels_0 fails when using NumPy 2.3.1, but passes when downgraded to NumPy 1.26.4.

To Reproduce

1. Create a conda env with:

• monai==1.5.0
• numpy==2.3.1
• torch==2.6.0+cu126
• torchaudio==2.6.0+cu126
• other related packages

2. Run python -m unittest tests.transforms.test_signal_remove_frequency
3. See test failure AssertionError: Not equal to tolerance rtol=1e-05, atol=0

Expected behavior Tests should pass consistently across supported NumPy versions.

Environment

================================
Printing system config...
================================
System: Linux
Linux version: Ubuntu 24.04.1 LTS
Platform: Linux-5.10.16.3-microsoft-standard-WSL2-x86_64-with-glibc2.39
Processor: x86_64
Machine: x86_64
Python version: 3.12.11
Process name: python
Command: ['python', '-c', 'import monai; monai.config.print_debug_info()']
Num physical CPUs: 6
Num logical CPUs: 6
Num usable CPUs: 6
CPU usage (%): [6.8, 2.4, 3.1, 3.9, 3.9, 92.2]
CPU freq. (MHz): 3696
Load avg. in last 1, 5, 15 mins (%): [1.3, 0.5, 0.1]
Disk usage (%): 12.5
Avg. sensor temp. (Celsius): UNKNOWN for given OS
Total physical memory (GB): 25.0
Available memory (GB): 22.1
Used memory (GB): 2.5

================================
Printing GPU config...
================================
Num GPUs: 1
Has CUDA: True
CUDA version: 12.6
cuDNN enabled: True
NVIDIA_TF32_OVERRIDE: None
TORCH_ALLOW_TF32_CUBLAS_OVERRIDE: None
cuDNN version: 90501
Current device: 0
Library compiled for CUDA architectures: ['sm_50', 'sm_60', 'sm_70', 'sm_75', 'sm_80', 'sm_86', 'sm_90']
GPU 0 Name: NVIDIA GeForce RTX 4060 Ti
GPU 0 Is integrated: False
GPU 0 Is multi GPU board: False
GPU 0 Multi processor count: 34
GPU 0 Total memory (GB): 16.0
GPU 0 CUDA capability (maj.min): 8.9

[ericspod]

Hi @IamTingTing thanks for point this out. I have created a PR which I think fixes the issue but I'll wait for all tests to pass to see if it works across versions. Please have a look to see if this fixes your issue.

[IamTingTing]

Hi @ericspod thanks for the fix! I’ve tested the PR on my environment with NumPy 2.3.1 and can confirm the unit test now passes.

[dzenanz]

Has that PR been merged now? Can this issue be closed?

[IamTingTing]

Hi @dzenanz ,thank you for following up. Since the PR is not yet merged, I think it would be better to keep this issue open until the changes are included in main.

[ericspod]

We should get #8537 through soon, we're sorting out a CI issue currently.