Why so many gunicorn worker processes?

Question

Django

workers = 4 threads = 1Tell me why if you specify and start the server in the gunicorn configs , the command pstree -ap|grep gunicornissues

|-gunicorn: maste,22988                                    
  |   |-gunicorn: worke,20372                                    
  |   |-gunicorn: worke,20373                                    
  |   |-gunicorn: worke,20374                                    
  |   `-gunicorn: worke,20375                                    
  |                   |-grep,20386 gunicorn

But after a load of requests:

|-gunicorn: maste,22988                                    
  |   |-gunicorn: worke,20372                                    
  |   |   |-{gunicorn: worke},20390
  |   |   `-{gunicorn: worke},20391
  |   |-gunicorn: worke,20373                                    
  |   |   |-{gunicorn: worke},20394
  |   |   `-{gunicorn: worke},20395
  |   |-gunicorn: worke,20375                                    
  |   |   |-{gunicorn: worke},20388
  |   |   `-{gunicorn: worke},20389
  |   `-gunicorn: worke,20396                                    
  |                   |-grep,20401 gunicorn

Inside the django application, I log pid and ppid, it turns out that each worker creates 2 workers when loaded. What is the logic behind this? What parameters set the ability to spawn processes by the workers themselves, and not by the master process?