Troubleshoot
Pipeline crash
If the coppafish pipeline is crashing, first read the error message. If there is a suggestion about how to fix the issue in the config, try changing the config variable and run the pipeline again. If the suggestion does not make sense to you, feel free to reach out to the developers for help or create an issue on GitHub!
Notebook will not open
A notebook file can be corrupted if a process is killed while the notebook is being re-saved. When this happens, an error like:
TypeError: byte indices must be integers or slices, not tuple
will occur when trying to load the notebook. To fix this issue, delete the corrupted notebook, rename the backup
notebook called notebook_backup.npz
to the original notebook name and continue from there.
Cannot open napari issues
If napari fails to open and you see an error such as
WARNING: composeAndFlush: makeCurrent() failed
when trying to open the Viewer or RegistrationViewer, here are a few suggestions that might fix the issue:
- In the conda environment, run
conda install -c conda-forge libstdcxx-ng
- In the conda environment, run
conda install -c conda-forge libffi
.
Filter image clip error
An error can occur when a filtered image clips off too many pixels when trying to save. This happens because the filter step will scale up every non-DAPI image by a common factor to improve precision. There are two options to deal with this issue:
- Reduce image clipping by lowering
scale_multiplier
below the default value found in thefilter
config (the default is found here). After this, delete thefilter
directory found in the tiles directory and thescale.txt
. Then, restart the pipeline. - Follow a "I don't care" strategy by increasing
percent_clip_error
above the default to allow for more clipped pixels. You can then restart the pipeline without deleting any files. If you wish to ignore warnings too, increasepercent_clip_warn
.
Memory crash at OMP
Try lowering subset_size_xy in the OMP config. This will cause OMP to compute on fewer pixels at time. It has a minimal effect on compute times, but can lower the RAM/VRAM usage. The default is found here.