Skip to content

Troubleshoot

Pipeline crash

If the coppafisher 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 developer(s) for help or create an issue on GitHub!

If you find any mysterious crash during filter or registration with an error message that traces back to the joblib python package, disable joblib entirely by adding

[filter]
num_cores = 1

for the filter stage and

[register]
flow_cores = 1

for the registration stage.

Find spots does not find enough spots

This is a common issue that happens. It is usually caused by the default automatic threshold parameters not fitting your particular dataset due to a difference in Signal-to-Noise Ratio. If your pipeline stops after find spots due to this reason, follow these steps:

  1. Open the find spots viewer, explained under the Detected Spots section.

  2. Here, you can tweak the multiplier and percentile and see how this affects the spot detections. You can choose to have a single multiplier and percentile shared for all images or you can choose different values for different channels.

  3. Once you are happy with the parameter tweaks, add the new values into the dataset's configuration file, e.g. 

    [find_spots]
auto_thresh_multipliers = 10
auto_thresh_percentiles = 25
    for the same values for all channels, or something like 
    [find_spots]
auto_thresh_multipliers = 10, 35, 5, 40
auto_thresh_percentiles = 25, 20, 25, 25
    for each sequence channel.

  4. Delete the notebook page find_spots, explained here.

  5. Run the pipeline again.

Memory crash at OMP

Try reducing subset_pixels in the OMP config. This will cause OMP to compute on fewer pixels at time. It has a minimal effect on compute times, but will lower the RAM/VRAM usage. By default, the subset pixel number is found on the fly based on the PC's hardware, you can see this by looking at the pipeline.log file and searching for subset_pixels.