ENHANCE-PET / MOOSE

MOOSE (Multi-organ objective segmentation) a data-centric AI solution that generates multilabel organ segmentations to facilitate systemic TB whole-person research.The pipeline is based on nn-UNet and has the capability to segment 120 unique tissue classes from a whole-body 18F-FDG PET/CT image.
https://enhance.pet
GNU General Public License v3.0
196 stars 28 forks source link

Moose could not start #137

Closed wolebode closed 2 months ago

wolebode commented 2 months ago

Hello, I tried installing moose following the instructions in the github. It says successfully installed. But when I try running moosez, it gives this error message:

OSError: [WinError 126] The specified module could not be found. Error loading "....\moose-env\lib\site-packages\torch\lib\fbgemm.dll" or one of its dependencies

I can confirm that this file "fbgemm.dll" is present in the folder. Please, any help will be appreciated.

image Capture

LalithShiyam commented 2 months ago

Hi @wolebode

This error is new for me as well. A search in the PyTorch forum pointed me towards this thread: https://discuss.pytorch.org/t/failed-to-import-pytorch-fbgemm-dll-or-one-of-its-dependencies-is-missing/201969

I think this is from torch and not really from moose.

The thread contains some solutions. Kindly have a look.

@mprires @Keyn34 tagging you for visibility.

Cheers, Lalith

mprires commented 2 months ago

Hi @wolebode

It appears you're on a Windows machine. Please note that PyTorch needs to be manually installed before installing MOOSE on Windows. We've encountered issues on another tool where skipping this step caused the same error for windows users. Could you confirm that you've installed PyTorch following the instructions at https://pytorch.org/get-started/locally/ before installing MOOSE?

Kind regards, Manel

wolebode commented 2 months ago

Thanks so much for your prompt response.

Yes, I installed PyTorch using the command "pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118" when that didn't work, I then started all over again and installed the CPU version using "pip3 install torch torchvision torchaudio"

Is that right?

Keyn34 commented 2 months ago

@wolebode, could you give us more information on the system you are trying MOOSE on? Which GPU, CPU, and Python Version are you using?

Also, I saw from the terminal that you tried to do moosez -d. Is this correct, or was the command trimmed due to the error?

Thanks in advance!

wolebode commented 2 months ago

Here are the details:

CPU - Intel(R) Core(TM) i9-14900K GPU - NVIDIA RTX 4000 Ada Generation Python - 3.10.0

About the command, it gives the same error whatever I type, even when I type "moosez -h" for help

wolebode commented 2 months ago

Thanks everyone. I have been able to resolve the issue. Invariably, it appeared that PyTorch was not properly installed. So, I installed CUDA NVIDIA, and the issue was fixed.

Thanks once again

LalithShiyam commented 2 months ago

@wolebode fantastic - keep us posted, if you are pleased with the results from MOOSE. And we are happy to help :)

wolebode commented 2 months ago

I will do. Thanks so much. You guys are awesome.

wolebode commented 2 months ago

Hello, it's me again: I'm trying to run moose on a CT dicom file, and it's giving this error: image I tried remaining all files to have CT1, CT2...CT86 but I don't know if there's anything else I need to do to make it compliant

LalithShiyam commented 2 months ago

Hi there, it runs in batch, can you send me the screenshot of your folder structure, it should be

All-cts |_ct_1 |_ct_2 |_ct_3

and the input is moosez -d <path_to_All-cts> -m clin_ct_organs

wolebode commented 2 months ago

image

LalithShiyam commented 2 months ago

Hi @wolebode, this folder should be nested inside another folder, and you should pass that folder as the input. Because moose runs in batches. Right now you are passing one of the subjects directly to moose.

wolebode commented 2 months ago

Thanks so much. It works!