In this document you will find instructions to
Jaxon is a container based service it uses docker-swarm and Nvidia GPUs for hardware acceleration. The Jaxon UI, runs on any one of the popular browsers, the UI connects to the running service over TCP port:31415. All Machine Learning (ML) operations are conducted via this UI.
Jaxon can run on a personal desktop or on a Virtual Machine Instance created in the public cloud. The recommended operating system and hardware specifications are provided in the table below.
| Operating System | Debian or Ubuntu LTS 20.04 |
|---|---|
| CPU Cores or vCPUs | 12-Cores minimum. |
| GPU | Nvidia GPU, with CUDA Version: 11.4 or higher |
| RAM | 60 Gig |
| Disk Capacity for Jaxon | 256 GB or more based on number of datasets and models. |
| Open source packages required on Jaxon | - Docker version 20 or higher |
The table above provides base specification to get started - Jaxon is a parallel compute service, it can take advantage of multiple GPUs and a larger bank of CPU Cores and RAM.
Each worker is assigned compute resource based on configuration specified in config.json file. All resource assignment occurs based on general pool available to the system.
Key components to consider are:
See configuration section below for adjusting resource assignments to better suite your workloads.
Ubuntu 20.04 or the equivalent Debian based operating system is recommended for Jaxon. CentOS or Red Hat Distribution may also be used but are not described here.
Jaxon is fully capable of running on a desktop server or a virtual-machine instance created in any one of the public-cloud. For illustration purposes we will use AWS.