How to batch submit 2D classification jobs

2D Classification
1

Hi, everyone! Does anyone know how to batch submit 2D classification jobs in CryoSAPAC? For example, I split a large particles dataset in several small datasets followed by creating 2D cls jobs for them. How to batch submit or create jobs in the GUI ?
Thanks very much!

2

How many small particle sets do you need 2D classified? Is the GUI constraint in your question “negotiable”?

3

Hi, wtempel! The large dataset contains 10 million particles which should be splited into 0.5 million per copy. I think other methods will be ok if works. Thanks!

4

This is a use case for cryosparc-tools.
You may:

  1. ensure prerequisites and install cryosparc-tools in a conda environment
  2. edit this script
#! /usr/bin/env python

# Requires: project and job ids of an existing Particle Set Tools job
# implementing the split of a particle set, specifying the split size

# Will queue 2d classification of subsets from splitting
# (including classification of the remainder)

from cryosparc.tools import CryoSPARC

# Instance info. Must edit.
license_id = "681437fb-d6ae-47b8-870b-b530c587da94"
master_host = "cryosparc.your.domain"
base_port = 39000
email = "your@email.address" # you usually user for login to the CryoSPARC UI
password = "toP_s4cret"
worker_lane_name = "default" # from "cryosparcm cli 'get_scheduler_lanes()'"

# Job settings. Must edit.
project_uid = "P147"
particle_set_tools_uid = "J33" # where you have already split your set

# Change job parameters below if needed

cs = CryoSPARC(license=license_id,
               host=master_host,
               base_port=base_port,
               email=email,
               password=password)

splits = {split.get("group_name") for split in cs.cli.get_job(
          project_uid, particle_set_tools_uid).get("output_results")}              

# 2D classification jobs will be created here
workspace = cs.create_workspace(project_uid, "many 2D class jobs")

for split in splits:
    class_2d_job = workspace.create_job("class_2D")
    class_2d_job.connect("particles", particle_set_tools_uid, split)
    class_2d_job.set_param("compute_use_ssd", "False")
    class_2d_job.set_param("class2D_K", 50)
    class_2d_job.queue(worker_lane_name)
  1. Save the script in a file, say many_2d.py
  2. activate the conda environment that you created for the installation of cryosparc-tools and run
    python many_2d.py
4 Likes