aydin fails to install on Windows

[nztao]

Describe the bug
Installing aydin via conda or pip on a Windows installation fails because of an incompatibility error with napari and pydantic.

To Reproduce
Steps to reproduce the behavior:

1. Run conda installation instructions from (https://github.com/royerlab/aydin).
2. Run aydin in the bash.

Expected behavior
A clear and concise description of what you expected to happen.
Expect aydin GUI, but instead the following error appears:

> [...]
>     raise AttributeError(f'module {module!r} has no attribute {name!r}')                                                                                                                                                                                                                                                                                  AttributeError: module 'pydantic.main' has no attribute 'ModelMetaclass'

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• Windows 10 Education Version 10.0.19045 Build 19045
• GPU: NVIDIA GeForce RTX 2080 Super - Cuda compilation tools, release 11.4, V11.4.100 Build cuda_11.4.r11.4/compiler.30188945_0

Additional context
Add any other context about the problem here.

Running pip install napari --upgrade resolves the issue.

[pmascalchi]

Having the same issue trying to run Aydin as standalone (in similar environment).
Any other possible fix found by anyone?
Thanks,

[IanWardell]

Similar problem on my end:
Desktop (please complete the following information):
Windows 11 Education Version
GPU: NVIDIA GeForce RTX 3090
powershell, conda, pip

Additional context
Ill start by saying the windows install exe works fine,its the CLI that wasn't working.

fixed with
pip install napari --upgrade

error:

(aydin_env) PS T:\Cam-Denoise> aydin info .\Images-To-Denoise\00001-capture.jpg Traceback (most recent call last): File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\runpy.py", line 197, in _run_module_as_main return _run_code(code, main_globals, None, File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\runpy.py", line 87, in _run_code exec(code, run_globals) File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\Scripts\aydin.exe\__main__.py", line 4, in <module> File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\__init__.py", line 1, in <module> from aydin.restoration.denoise.noise2selffgr import noise2self_fgr # noqa: F401 File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\restoration\denoise\noise2selffgr.py", line 8, in <module> from aydin.features.standard_features import StandardFeatureGenerator File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\features\standard_features.py", line 5, in <module> from aydin.features.groups.lowpass import LowPassFeatures File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\features\groups\lowpass.py", line 5, in <module> from aydin.it.classic_denoisers.butterworth import denoise_butterworth File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\it\classic_denoisers\butterworth.py", line 13, in <module> from aydin.util.j_invariance.j_invariance import calibrate_denoiser File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\util\j_invariance\j_invariance.py", line 7, in <module> from aydin.util.j_invariance.losses import mean_squared_error, mean_absolute_error File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\aydin\util\j_invariance\losses.py", line 3, in <module> from skimage.metrics import structural_similarity File "C:\Users\GOOD-USER\anaconda3\envs\aydin_env\lib\site-packages\skimage\__init__.py", line 121, in <module> from ._shared import geometry File "skimage\_shared\geometry.pyx", line 1, in init skimage._shared.geometry ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

[manerotoni]

I can confirm to have issues when installing on Windows10.

Traceback (most recent call last):
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\runpy.py", line 197, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\Scripts\aydin.exe\__main__.py", line 4, in <module>
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\__init__.py", line 1, in <module>
    from aydin.restoration.denoise.noise2selffgr import noise2self_fgr  # noqa: F401
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\restoration\denoise\noise2selffgr.py", line 8, in <module>
    from aydin.features.standard_features import StandardFeatureGenerator
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\features\standard_features.py", line 5, in <module>
    from aydin.features.groups.lowpass import LowPassFeatures
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\features\groups\lowpass.py", line 5, in <module>
    from aydin.it.classic_denoisers.butterworth import denoise_butterworth
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\it\classic_denoisers\butterworth.py", line 13, in <module>
    from aydin.util.j_invariance.j_invariance import calibrate_denoiser
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\util\j_invariance\j_invariance.py", line 7, in <module>
    from aydin.util.j_invariance.losses import mean_squared_error, mean_absolute_error
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\aydin\util\j_invariance\losses.py", line 3, in <module>
    from skimage.metrics import structural_similarity
  File "C:\Users\apoliti\Miniconda3\envs\aydin_env\lib\site-packages\skimage\__init__.py", line 121, in <module>
    from ._shared import geometry
  File "skimage\_shared\geometry.pyx", line 1, in init skimage._shared.geometry
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

The command pip install napari --upgrade fixes the problem (why??) although the install mention compatibilities issues

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
aydin 0.1.15 requires imageio==2.21.1, but you have imageio 2.35.1 which is incompatible.
aydin 0.1.15 requires napari==0.4.15, but you have napari 0.5.2 which is incompatible.
aydin 0.1.15 requires scikit-image==0.18.3, but you have scikit-image 0.24.0 which is incompatible.

[Downtown-Case]

I have this same issue on (arch) linux:

ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

Installing with conda fails, installing with plain uv and python 3.9 fails the same way.

If I download the studio zip, the GUI fails to start with a segfault. If I try denoising an image, it does start (though only with the studio version), then:

/home/alpha/AI/aydin/aydin/aydin denoise /home/alpha/FastStorage/Data/021.jpg --output-folder /home/alpha/FastStorage/Data/denoise/                                            (aydin_env) 
WARNING:tensorflow:AutoGraph is not available in this environment: functions lack code information. This is typical of some environments like the interactive Python shell. See https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/autograph/g3doc/reference/limitations.md#access-to-source-code for more information.
Unable to read version number
QFont::fromString: Invalid description 'Atkinson Hyperlegible Next,11,-1,5,500,0,0,0,0,0,0,0,0,0,0,1,Medium'
QFont::fromString: Invalid description 'Atkinson Hyperlegible Mono,11,-1,5,500,0,0,0,0,0,0,0,0,0,0,1,Medium'
QFont::fromString: Invalid description 'Atkinson Hyperlegible Next,11,-1,5,400,0,0,0,0,0,0,0,0,0,0,1'
QFont::fromString: Invalid description 'Atkinson Hyperlegible Next,11,-1,5,500,0,0,0,0,0,0,0,0,0,0,1,Medium'
build/linux/runtimehooks/hook-splash.py:15: MatplotlibDeprecationWarning: 
The set_window_title function was deprecated in Matplotlib 3.4 and will be removed two minor releases later. Use manager.set_window_title or GUI-specific methods instead.
qt.qpa.wayland: Failed to load client buffer integration: "wayland-egl"
qt.qpa.wayland: Available client buffer integrations: ()
qt.qpa.wayland: No shell integration named "xdg-shell" found
qt.qpa.wayland: No shell integration named "xdg-shell-v6" found
qt.qpa.wayland: No shell integration named "wl-shell" found
qt.qpa.wayland: No shell integration named "ivi-shell" found
qt.qpa.wayland: Loading shell integration failed.
qt.qpa.wayland: Attempted to load the following shells ("xdg-shell", "xdg-shell-v6", "wl-shell", "ivi-shell")
qt.qpa.wayland: Wayland does not support QWindow::requestActivate()
qt.qpa.wayland: Wayland does not support QWindow::requestActivate()
├╗ Reading image file at: /home/alpha/FastStorage/Data/LoveBites-S01E01-021.jpg
│├ Reading file /home/alpha/FastStorage/Data/LoveBites-S01E01-021.jpg as PNG file
│├ Metadata:  is_folder=False, ext=jpg, axes=YXC, shape=(720, 1280, 3), batch_axes=(False, False, False), channel_axes=(False, False, True), dtype=uint8, format=jpg 
│┴« 9.82 milliseconds
│
├ Image file format is not supported, will be writing result as tif
├╗ Noise2Self train is starting...
│├ Features will be computed using dtype: <class 'numpy.float32'>
│├╗ CB Regressor
││├ patience: 32
││├ gpu: True
││┴« 11.44 microseconds
││
│├╗ FGR image translator
││├ balance training data: False
││┴« 6.44 microseconds
││
│├ Instanciating: RangeTransform (mode=minmax, percentile=None, leave_as_float=False, clip=True )
│├ Instanciating: PaddingTransform (pad_width=3, min_length_to_pad=8, mode=reflect)
│├ Instanciating: VarianceStabilisationTransform (mode=anscomb)
│├╗ Learning to translate from image of dimensions (720, 1280, 3) to (720, 1280, 3), batch_axes=(False, False, False), channel_axes=(False, False, True), jinv=None.
││├ Running garbage collector...
││├ Training is self-supervised.
││├╗ transform preprocess:
│││├ applying transform: VarianceStabilisationTransform (mode=anscomb)
│││├╗ Stabilising variance (anscomb) for array of shape: (720, 1280, 3) and dtype: uint8
││││┴« 463.25 milliseconds
││││
│││├ applying transform: RangeTransform (mode=minmax, percentile=None, leave_as_float=False, clip=True )
│││├╗ Normalizing value range (minmax) for array of shape: (720, 1280, 3) and dtype: float32
││││├╗ Calibrating array using minmax method
│││││├ Range for normalisation: [1.2247449159622192, 31.960912704467773]
│││││┴« 443.46 microseconds
│││││
││││┴« 6.30 milliseconds
││││
│││├ applying transform: PaddingTransform (pad_width=3, min_length_to_pad=8, mode=reflect)
│││├╗ Padding array of shape: (720, 1280, 3) with 3 voxels and mode: reflect:
││││├ Effective padding widths: ((3, 3), (3, 3), (0, 0))
││││┴« 595.33 microseconds
││││
│││┴« 470.35 milliseconds
│││
││├ Automatic discovery of noise autocorrelation and specification of N2S blind-spots activated!
││├╗ Cropping image of size: (636, 1126) with at most 1000000 voxels and mode contrast
│││├╗ Cast and normalise image...
││││┴« 1.91 microseconds
││││
│││├╗ Apply cropping filter to image of shape: (636, 1126)
││││┴« 3.19 milliseconds
││││
│││┴« 3.23 milliseconds
│││
││├ Blind spots: [(-1, 0), (0, 0), (1, 0)]
││├ Autocorrelogram unique values in decreasing order: [+1.    ,+0.7469,+0.7465,+0.7305,+0.7303]
││├╗ Training image translator from image of shape (1, 3, 726, 1286) to image of shape (1, 3, 726, 1286):
│││├ Image has: 2800908 voxels, at most: 40000000 voxels will be used for training or validation.
│││├ Given train ratio is: 1, max_voxels induced keep-ratio is: 14.28108313446925
│││├ Data histogram balancer is: inactive
│││├ Effective keep-ratio is: 1.0
│││├ Favouring bright pixels: no
│││├ Using contiguous batches of length: 16 
│││├ Calibrating balancer...
│││├ Balancer: no calibration needed 
│││├╗ Determine tilling strategy:
││││├ image shape             = (1, 3, 726, 1286)
││││├ max_voxels_per_tile     = 452984832
││││├╗ Estimating the amount of memory needed to store feature arrays:
│││││├╗ Computing features for image of shape (1, 1, 4, 4):
││││││├ exclude_center_feature = True
││││││├ exclude_center_value   = False
││││││├ excluded_voxels        = [(-1, 0), (1, 0)]
││││││├ spatial_feature_scale     = None
││││││├ spatial_feature_offset    = None
││││││├╗ Computing features
│││││││├ Number of duplicate features: 0
│││││││├╗ Creating feature array for image of shape: (1, 1, 4, 4)
││││││││├╗ Array of shape: (54, 1, 4, 4) and dtype: <class 'numpy.float32'> requested
│││││││││├ Array requested will be 0.003456 MB.
│││││││││├ There is 32183 MB of physical memory
│││││││││├ There is 36371 MB of swap memory
│││││││││├ There is 68554 MB of total memory
│││││││││├ There is enough physical+swap memory -- we do not need to use a mem mapped array or zarr-backed array.
│││││││││┴« 325.68 microseconds
│││││││││
││││││││┴« 347.85 microseconds
││││││││
│││││││├╗ Computing features for batch: 1/1
││││││││├╗ Computing features for channel: 1/1
│││││││││├ Excluding center value for channel: False
│││││││││├ Image slice: (0, 0, slice(None, None, None), slice(None, None, None))
│││││││││├ Feature slice for batch and channel: (slice(None, None, None), 0, slice(None, None, None), slice(None, None, None))
│││││││││├ Computing feature <aydin.features.groups.uniform.UniformFeatures object at 0x7fe168a7b490>, 
│││││││││├ Pre-computing uniform filter of size: (2, 3)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (2, 3) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (3, 2)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (3, 2) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (3, 5)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (3, 5) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (5, 3)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (5, 3) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (4, 7)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (4, 7) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (7, 4)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (7, 4) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (3, 3)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (3, 3) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (7, 7)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (7, 7) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (15, 15)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (15, 15) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (31, 31)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (31, 31) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (63, 63)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (63, 63) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (127, 127)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (127, 127) with parallel scipy
│││││││││├ Pre-computing uniform filter of size: (255, 255)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (255, 255) with Numba
│││││││││├ Pre-computing uniform filter of size: (511, 511)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (511, 511) with Numba
│││││││││├ Pre-computing uniform filter of size: (1023, 1023)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (1023, 1023) with Numba
│││││││││├ Pre-computing uniform filter of size: (2047, 2047)
│││││││││├ Image too small, CUDA not needed!
│││││││││├ Computed filter of size: (2047, 2047) with Numba
│││││││││├ Uniform feature: 0, description: ((0, 0), (0, 1), (1, 1), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 1, description: ((0, 0), (1, 0), (1, 1), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 2, description: ((0, 0), (1, 1), (1, 0), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 3, description: ((0, 0), (1, 1), (0, 1), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 4, description: ((0, 0), (0, 2), (2, 2), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 5, description: ((0, 0), (2, 0), (2, 2), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 6, description: ((0, 0), (2, 2), (2, 0), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 7, description: ((0, 0), (2, 2), (0, 2), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 8, description: ((0, 0), (0, 3), (3, 3), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 9, description: ((0, 0), (3, 0), (3, 3), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 10, description: ((0, 0), (3, 3), (3, 0), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 11, description: ((0, 0), (3, 3), (0, 3), '#l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 12, description: ((-3, 0), (1, 1), (1, 1), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 13, description: ((0, -3), (1, 1), (1, 1), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 14, description: ((0, 0), (1, 1), (1, 1), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 15, description: ((0, 3), (1, 1), (1, 1), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 16, description: ((3, 0), (1, 1), (1, 1), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 17, description: ((-7, 0), (3, 3), (3, 3), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 18, description: ((0, -7), (3, 3), (3, 3), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 19, description: ((0, 0), (3, 3), (3, 3), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 20, description: ((0, 7), (3, 3), (3, 3), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 21, description: ((7, 0), (3, 3), (3, 3), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 22, description: ((-15, 0), (7, 7), (7, 7), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 23, description: ((0, -15), (7, 7), (7, 7), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 24, description: ((0, 0), (7, 7), (7, 7), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 25, description: ((0, 15), (7, 7), (7, 7), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 26, description: ((15, 0), (7, 7), (7, 7), 'l2'), excluded_voxels=[]
│││││││││├ Uniform feature: 27, description: ((-31, 0), (15, 15), (15, 15), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 28, description: ((0, -31), (15, 15), (15, 15), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 29, description: ((0, 31), (15, 15), (15, 15), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 30, description: ((31, 0), (15, 15), (15, 15), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 31, description: ((-63, 0), (31, 31), (31, 31), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 32, description: ((0, -63), (31, 31), (31, 31), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 33, description: ((0, 63), (31, 31), (31, 31), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 34, description: ((63, 0), (31, 31), (31, 31), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 35, description: ((-127, 0), (63, 63), (63, 63), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 36, description: ((0, -127), (63, 63), (63, 63), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 37, description: ((0, 127), (63, 63), (63, 63), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 38, description: ((127, 0), (63, 63), (63, 63), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 39, description: ((-255, 0), (127, 127), (127, 127), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 40, description: ((0, -255), (127, 127), (127, 127), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 41, description: ((0, 255), (127, 127), (127, 127), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 42, description: ((255, 0), (127, 127), (127, 127), 'l1nc'), excluded_voxels=[]
│││││││││├ Uniform feature: 43, description: ((0, 0), (255, 255), (255, 255), 'l1oc'), excluded_voxels=[]
│││││││││├ Uniform feature: 44, description: ((0, 0), (511, 511), (511, 511), 'l1oc'), excluded_voxels=[]
│││││││││├ Uniform feature: 45, description: ((0, 0), (1023, 1023), (1023, 1023), 'l1oc'), excluded_voxels=[]
│││││││││├ Computing feature <aydin.features.groups.lowpass.LowPassFeatures object at 0x7fe168a7b700>, 
│││││││││├ Correlative feature: 0 of shape=(5, 5), excluded_voxels=[]
│││││││││├ Correlative feature: 1 of shape=(7, 7), excluded_voxels=[]
│││││││││├ Correlative feature: 2 of shape=(7, 7), excluded_voxels=[]
│││││││││├ Correlative feature: 3 of shape=(7, 7), excluded_voxels=[]
│││││││││├ Correlative feature: 4 of shape=(7, 7), excluded_voxels=[]
│││││││││├ Correlative feature: 5 of shape=(9, 9), excluded_voxels=[]
│││││││││├ Correlative feature: 6 of shape=(9, 9), excluded_voxels=[]
│││││││││├ Correlative feature: 7 of shape=(9, 9), excluded_voxels=[]
│││││││││┴« 1.43 seconds
│││││││││
││││││││┴« 1.43 seconds
││││││││
│││││││┴« 3.32 seconds
│││││││
││││││┴« 3.32 seconds
││││││
│││││┴« 3.32 seconds
│││││
││││├ Estimated amount of memory needed: 786494966.4
││││├ Available memory (we reserve 10% for 'comfort'): 28964881612.8
││││├ How much do we need to tile because of memory? : 0.02715339827428939 times.
││││├ How much do we need to tile because of the limit on the number of voxels per tile? : 0.006183226903279622 times.
││││├ How much do we need to tile because of the suggested tile size? : 0.0 times.
││││├ Desired split factor: 1
││││├ Tilling strategy is: (1, 1, 1, 1)
││││├ The estimated tile shape is: (726, 1286)
││││┴« 3.32 seconds
││││
│││├ Tilling strategy (just batches): (1, 1, 1, 1)
│││├ Margins for tiles: (0, 0, 0, 0) .
│││├ Number of tiles (slices): 1
│││├╗ Current tile: 0/1, slice: (slice(0, 1, 1), slice(0, 3, 1), slice(0, 726, 1), slice(0, 1286, 1)) 
││││├╗ Computing features for image of shape (1, 3, 726, 1286):
│││││├ exclude_center_feature = True
│││││├ exclude_center_value   = True
│││││├ excluded_voxels        = [(-1, 0), (1, 0)]
│││││├ spatial_feature_scale     = (0.0013774104683195593, 0.0007776049766718507)
│││││├ spatial_feature_offset    = (0, 0)
│││││├╗ Computing features
││││││├ Number of duplicate features: 0
││││││├╗ Creating feature array for image of shape: (1, 3, 726, 1286)
│││││││├╗ Array of shape: (162, 1, 726, 1286) and dtype: <class 'numpy.float32'> requested
││││││││├ Array requested will be 604.996128 MB.
││││││││├ There is 32183 MB of physical memory
││││││││├ There is 36371 MB of swap memory
││││││││├ There is 68554 MB of total memory
││││││││├ There is enough physical+swap memory -- we do not need to use a mem mapped array or zarr-backed array.
││││││││┴« 324.01 microseconds
││││││││
│││││││┴« 345.47 microseconds
│││││││
││││││├╗ Computing features for batch: 1/1
│││││││├╗ Computing features for channel: 1/3
││││││││├ Excluding center value for channel: True
││││││││├ Image slice: (0, 0, slice(None, None, None), slice(None, None, None))
││││││││├ Feature slice for batch and channel: (slice(None, None, None), 0, slice(None, None, None), slice(None, None, None))
││││││││├ Computing feature <aydin.features.groups.uniform.UniformFeatures object at 0x7fe168a7b490>, 
││││││││├ Pre-computing uniform filter of size: (2, 3)
fish: Job 1, '/home/alpha/AI/aydin/aydin/aydi…' terminated by signal SIGSEGV (Address boundary error)