Skip to content
forked from corresive/dop-q

Lightweight docker priority queue implementation for running a multi-gpu deep learning system in a job processing mode. The system keeps a history of executions per user to calculate penalties and allow a fair amount of GPU processing power for everyone. The main purpose of this software is to run an arbitrary amount of jobs of multiple users on…

Notifications You must be signed in to change notification settings

IljaManakov/dop-q

 
 

Repository files navigation

DoP-Q

Queue for docker to run projects on a multi-gpu machine

History:

  • 12.10.2017: Initial commit.
  • 16.10.2017: Changed to python 2.7
  • 31.01.2018: Refactored queue from the ground up. Introduced modules builder.py, container_handler.py, gpu_handler.py and helper_process.py

HowTo get started:

Just download run_python_script in examples/simple and zip the content of the folder. Name it to "build_[SOME_NAME]_[YOUR_NAME].zip", where [SOME_NAME] is some name you may freely choose and where [YOUR_NAME] represents your username. Copy it to the container.path directory of the queue and it will be built and run automatically. Please not that [YOUR_NAME] must be authorized to run docker files on the machine. Please speak to some administrator of the machine (Ilja Manakov, Markus Rohm).

Update History:

  • 15.04.2019: Decided to move in new interface for better flexibility and introduce server-client communication.
  • 21.06.2019: Beta version of Pyqt4 interface has been integrated to the backend.
  • 18.07.2019: Decided to separate the docker priority queue and provider process completely from the interface. Agreed to implement the system in MVC fashion with client totally separated from the server.
pyqt4_ui
New Interface of DoPQ [Beta Version].

About

Lightweight docker priority queue implementation for running a multi-gpu deep learning system in a job processing mode. The system keeps a history of executions per user to calculate penalties and allow a fair amount of GPU processing power for everyone. The main purpose of this software is to run an arbitrary amount of jobs of multiple users on…

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 99.9%
  • Shell 0.1%