TuragaLab / DECODE

This is the official implementation of our publication "Deep learning enables fast and dense single-molecule localization with high accuracy" (Nature Methods)
GNU General Public License v3.0
95 stars 26 forks source link

Package installation is broken in Ubuntu 20.24 #237

Open mmarinriera opened 2 months ago

mmarinriera commented 2 months ago

Hi,

I tried to install the package following the installation instructions on a machine with Ubuntu22.04, with a CUDA capable GPU

I followed the installation steps in the documentation:

conda config --set channel_priority strict
conda create -n decode_env -c turagalab -c pytorch -c conda-forge decode=0.10.2 cudatoolkit=11.3 jupyterlab ipykernel
conda activate decode_env

Installation is successful, however, when I try to import decode with Python, I get this error message:

Python 3.8.19 | packaged by conda-forge | (default, Mar 20 2024, 12:47:35)
[GCC 12.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import decode
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/miquel/miniconda3/envs/decode_env/lib/python3.8/site-packages/decode/__init__.py", line 14, in <module>
    import decode.evaluation
  File "/home/miquel/miniconda3/envs/decode_env/lib/python3.8/site-packages/decode/evaluation/__init__.py", line 1, in <module>
    import decode.evaluation.evaluation
  File "/home/miquel/miniconda3/envs/decode_env/lib/python3.8/site-packages/decode/evaluation/evaluation.py", line 8, in <module>
    import torch
  File "/home/miquel/miniconda3/envs/decode_env/lib/python3.8/site-packages/torch/__init__.py", line 202, in <module>
    from torch._C import *  # noqa: F403
ImportError: /home/miquel/miniconda3/envs/decode_env/lib/python3.8/site-packages/torch/lib/libtorch_cpu.so: undefined symbol: iJIT_NotifyEvent
>>>

Regarding this error message, it has already been reported here, and it seems the solution is to downgrade the mkl package.

conda install -c conda-forge mkl==2024.0

That seemed to solve the issue with decode as well, but I don't know if it will break things at a lower level.

Hope this helps!