Patch motion correction misaligned address

When I run patch motion correction, I always get the error as shown in the figure, only 1000 of 4000 pictures complete the calculation

Welcome to the forum @zhe. Please can you tell us more about your case:

  • movie format (movie_blob/shape) from the Overview tab of Job J1
  • cryoSPARC worker info:
    • cryoSPARC version and patch level
    • gpu models and driver version (from nvidia-smi)
    • linux kernel version (from uname -a)
    • cpu version (lscpu | grep name)
    • worker cuda version (grep CUDA /opt/cryosparc/cryosparc_worker/
    • worker pycuda version:
eval $(/opt/cryosparc/cryosparc_worker/bin/cryosparcw env)
python -c "import pycuda.driver; print(pycuda.driver.get_version())"

Thank you very much for your reply. The information I found is as follows. thank you very much. Since I can only attach one picture at a time, I reply all the pictures in multiple pieces of information at the back

@zhe The import output you posted shows a warning near the top. Inspection of the entire output (“Show from top”) may reveal an underlying issue that needs to be resolved.

Yes, there is a warning. But this is because one photo failed to be imported. Here is the complete screenshot. I’m very sorry for the trouble to your work.

@zhe Is there any indication why the import failed?
On the other hand, other users have reported CUDA-related errors on kernel v3-based systems. The specific errors were different, and I am not sure the proposed interventions would solve the motion correction error you are experiencing.
How much RAM (free -g) does your system have?

As this picture shows, my friend helped me free up memory space. I’m a biology phD myself and not very computer savvy. Thank you very much for your help

@zhe What happens when you connect the movies that were marked as incomplete in the old patch motion job as input to a new patch motion job that is otherwise identical to the old job? Do any of those previously “incomplete” movies complete in the new job?

Thank you very much for your long and patient answer. After several attempts, the two GPUs (0,1) of our computer should be the failure of GPU0, and I will not have any failure when I only use GPU1