Troubleshooting

Napari issues

Black screen

napari opens to a black screen with the following error messages:

WARNING: QOpenGLWidget: Failed to create context
parallel.py (371): The TBB threading layer requires TBB version 2021 update 6 or later i.e., TBB_INTERFACE_VERSION >= 12060. Found TBB_INTERFACE_VERSION = 12050. The TBB threading layer is disabled.
WARNING: QOpenGLWidget: Failed to create context
WARNING: composeAndFlush: QOpenGLContext creation failed
WARNING: composeAndFlush: makeCurrent() failed

A potential fix is to create a symbolic link to libraries not found by napari in your python environment (here a conda environment named celldetective):

$ cd ~/anaconda3/envs/celldetective/lib
$ mkdir backup
$ mv libstd* backup
$ cp /usr/lib/x86_64-linux-gnu/libstdc++.so.6  ./
$ ln -s libstdc++.so.6 libstdc++.so
$ ln -s libstdc++.so.6 libstdc++.so.6.0.19

Deep learning libraries

Pytorch

On older hardware, Pytorch may yield the following error:

Pytorch: [W NNPACK.cpp:64] Could not initialize NNPACK! Reason: Unsupported hardware.

A potential fix is to install Pytorch through mamba:

$ mamba remove pytorch
$ mamba install pytorch

The following error (when training a model):

usr/local/lib/python3.10/dist-packages/torch/lib/../../nvidia/cusparse/lib/libcusparse.so.12: undefined symbol: __nvJitLinkAddData_12_1, version libnvJitLink.so.12

can be fixed by reinstalling torchvision and torchaudio:

$ python -m pip uninstall torch torchvision torchaudio
$ python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121

Tensorflow

On older hardware, Tensorflow may yield the following error:

Tensorflow: Illegal instruction (core dumped)
The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine.

A potential fix is to install Tensorflow through mamba:

$ conda remove tensorflow
$ mamba remove tensorflow
$ mamba install tensorflow

StarDist

When training a StarDist model on an older CPU the following error can be triggered:

pyopencl._cl.LogicError: clGetPlatformIDs failed: PLATFORM_NOT_FOUND_KHR

Try to install the missing pocl library as:

$ pip install pocl-binary-distribution

Multithreading

Setting too many threads for the device you are using may lead to jobs aborting silently. This has been observed in the segmentation jobs. If you notice that whole label frames are missing, try to decrease the number of threads in the settings.

Note

In general, avoid combining both the use of GPU and multithreading as it may lead to memory issues.