Requesting Tips for managing data between binned particles and unbinned particles

Hi Everyone,

I am a novice in cryosparc.
Now I am trying handling a big data set.
My trial was using 1000 micrographs to creat templates for picking. Then using templates, I picked almost 4 millions particles for initial 2D classification.
But most of further processing was so painful. The particle data was too big, couldn’t process by 2T SSD scratch disk. One round of 2D classification with regular HDD took an whole day.
After several days later, I learned about “Downsample” from here. Then I could use my scratch disk. But sometime, I still am getting lacks of GPU memory issues when I gave more options in 2D classifications.

Here are my questions,

  1. I heard I can extract particles with a binned sized 2X or 3X. Does it help for giving higher
    batchsizes, Max force off related with GPU memory issues, especially 2D classification
    speed up?
  2. After finishing 2D classifications or 3D reconstruction and refinements, Is there a way to
    recover selected binned particles to original size?

I appreciate all your advice!


For 2D classification, you can run multiple jobs and pool the particles later. For example if you have 1 000 000 particles, you can run 4 jobs with 250 000 particles.