Raptor Talos II Secure Workstation Setup Guide

This repository provides a comprehensive guide for setting up a Raptor Talos II Secure Workstation for AI, machine learning and robotics software development.

This system's motherboard has a dual-socket configuration and can support 4, 8, 18 and 22-core IBM Power9 CPUs.

Installation checklist

Operating system install

Install Ubuntu-18.04 LTS on a Raptor Talos II Power9 system:

• Ubuntu-18.04 LTS for ppc64el, server, ubuntu-desktop.

• NVIDIA proprietary graphics driver v418.67 for ppc64el, tested with the following GPUs:

  • Tesla V100 with 16GB HBM2
  • Titan-V with 12GB HBM2
  • Quadro RTX 8000 with 48GB GDDR6
  • GeForce RTX 2070 with 8GB GDDR6
  • Quadro P6000 with 24GB GDDR5

Graphics and acceleration libraries.

Install NVIDIA graphics and acceleration libraries:

• NVIDIA proprietary graphics driver 418.87
• CUDA-10.1
• cuDNN-7.6.2.24
• TensorRT-5.1.5.0
• NCCL-2.4.7

Remote desktop access.

Compile libjpeg-turbo, VirtualGL and TurboVNC from sources:

• libjpeg-turbo v2.0.2, compiled from source, branch: 2.0.2, commit id: a4aa30d9a080bbc50421285049e4379dcaf8a669
• VirtualGL v2.6.3, compiled from source, branch: master, commit id: b74c7b86f21703b201682e2078e1188fc46c3b1f
• TurboVNC v2.2.2, compiled from source, branch: master, commit id: 8cf390455d62dfd50595f9edfc53bc1349a4c481

Choose any one of the following options to configuring remote desktop access:

• Configure Ubuntu-18.04 remote desktop access using VirtualGL and TurboVNC using gdm3 display manager (preferred).
• Configure Ubuntu-18.04 remote desktop access using VirtualGL and TurboVNC using lightdm display manager.

Common development libraries.

• GoogleTest, built from source, branch: master, commit id: 90a443f9c2437ca8a682a1ac625eba64e1d74a8a.

Common development tools.

• CLion-2019.2 with OpenJDK 11.0.4 2019-07-16.
• PyCharm-2019.2 with OpenJDK 11.0.4 2019-07-16.

DevOps.

• docker-ce on ppc64el.
• nvidia-docker2 on ppc64el, verify nvidia gpu access from docker container on ppc64el.
• kubernetes v1.14.3 on ppc64el.
• grafana dashboard.
• kubeflow.

Game engines.

• UE4.22.3 port to ppc64le.
• UE4.23.1 port to ppc64le.

Machine learning libraries.

Machine learning libraries:

• TensorFlow-2.0.0-beta1, compiled from source, branch: 2.0.0. commit id: ``
• PyTorch-1.2.0, compiled from source, branch: v1.2.0. commit id: ``

Robotics libraries.

• ROS1 Melodic Morena, compile from source for ppc64el.
• ROS2 Dashing Diademata, compile from source for ppc64el.

Robotics simulation environments and software development kits.

• OpenAI Gym
• NVIDIA Isaac SDK, UE4.22.3 sim.