Experiment Scheduler

The web-based scheduler tool to allow

1. multiple people share GPU workstations and
2. see the status, stdout and stderr of exeperiments
3. control experiments (start, stop).

Features

This sheduler is designed to miximize the usage of GPU workstation and provide fair and even computation resources to people.

Environment encapsulation

To allow everyone have their own environment for their experiments without interferring other poeple's environment, this scheduler will use docker containers to encapsulate each experiments.

Dependencies

• Docker
• Python3
• At least one GPU
• LDAP
• LDAP authentication server

Installation

# clone the scheduler from github
git clone https://github.com/yoshihikoueno/ExperimentScheduler
cd ExperimentScheduler

# [optional but recommended] prepare virtual environment for python packages
python3 -m venv env
source env/bin/activate

# install required packages
pip3 install -r requirements.txt

It'll be also required to setup authentication stuff. Please refer to external guide for setting up LDAP authentication server and client.

Usage

Example

python3 main.py  --logdir /tmp --public

options

python main.py --help

# usage: main.py [-h] [--logdir LOGDIR] [--config CONFIG] [--public]
#                [--port PORT]
# 
# optional arguments:
#   -h, --help       show this help message and exit
#   --logdir LOGDIR  Directory where all log files should be stored
#   --config CONFIG  Config file for the scheduler
#   --public         Whether to publish web server to the local network
#   --port PORT      The port to use for the web interface. Defaults to 5000

Screenshot

NOTE

This scheduler is developed as an internal tool to make it easier to schedule experiments on a server, so the documentation and implementation may not be appropriate for general usage outside our environment.