This guide will walk you through re-installing Ubuntu 18.04 on a Lambda Quad.
You’ll need:
- 1x USB Stick (8GB+)
- A Lambda Quad GPU Workstation
Step 1 - Install Ubuntu 18.04
1.1 Create a Ubuntu 18.04 installation medium by following this guide:
1.2 Plug your monitor into the HDMI slot on the first (Top) GPU.
1.3 Insert the USB into your computer
1.4 Restart your computer and hit [F2] to enter into the boot menu
1.5 Hit [F8] to choose a boot medium, then select UEFI: <VENDOR>, Partition 1
Note that mine says “UEFI: SanDisk, Partition 1”, yours might say something like “UEFI: Patriot, Partition 1” or something else depending on the vendor of your USB stick.
1.6 Use the down arrows to choose Install Ubuntu, then hit [E]
1.7 Edit the linux line by inserting nouveau.modeset=0 like below:
Before:
After:
Closeup:
1.8 Hit [F10] to start the installation
NOTE: you might see an error message like this, don’t worry, it will go away shortly and this won’t prevent you from installing Ubuntu.
Follow these visual instructions to install Ubuntu 18.04
1.9 Continue
1.10 Normal Installation > Continue
1.11 Erase Disk and Install Ubuntu > Continue
1.12 Select Installation Medium
Most Lambda Quads use a 2TB SSD for the OS Installation, choose that SSD and then hit Install Now.
1.13 Confirm that you’d like to wipe your disk to install Ubuntu
WARNING: this step will completely erase your disk drive and will cause irrecoverable data loss, do not hit Continue unless you are fine losing everything on this disk.
1.14 Choose your Location
1.15 Create your default user
1.16 Wait for installation to complete (takes some time)
Yay! So far so good. Click Restart Now
1.17 Remove the USB drive
Remove the USB drive and then reboot the system.
Remove your USB drive.
Hit [ENTER].
Note that you may encounter a non responsive black screen like I did after hitting [ENTER]. In that case, simply do a hard reboot by holding down the power button for a few seconds and then clicking it on again as demonstrated below.
1.17a Hard power cycle your machine.
Hold down the power button for a few seconds until all of the lights go off.
Tap the power button again to power it back on.
You should be greeted by the boot logo.
Great! Time to move on to installing the default Deep Learning software with Lambda Stack
Step 2 - Install Lambda Stack
After installation you’ll want to install Lambda Stack. The latest up-to-date instructions are always available here: Lambda Stack: an AI software stack that's always up-to-date.
2.1 Log into your new account
2.2 Open up a terminal
Click Show Applications button on the bottom right
Search for term
Click the Terminal application
Here is your Terminal window:
2.3 Get the Lambda Stack installation instructions
You can open FireFox and search Google for “Lambda Stack” or go to this link:
2.4 Copy the installation instructions
Highlight the instructions, right click, and then click “Copy”.
This is a snipped copied from the official Lambda Stack installation page as of September 7th, 2018
LAMBDA_REPO=$(mktemp) && \
wget -O${LAMBDA_REPO} https://lambdalabs.com/static/misc/lambda-stack-repo.deb && \
sudo dpkg -i ${LAMBDA_REPO} && rm -f ${LAMBDA_REPO} && \
sudo apt-get update && sudo apt-get install -y lambda-stack-cuda
2.5 Paste the instructions into the terminal
Right click and select “Paste”. Then hit [ENTER]
Paste
Hit [ENTER]
2.6 Type in your sudo password and hit [ENTER]
2.7 Wait a long time, this takes about 15-20 minutes, watch this NIPS talk in the mean time
It’s a classic on Differentiable Neural Computers by Alex Graves:
2.8 Read and Accept the CuDNN Licensing Agreement
Halfway through the installation this dialog menu will pop up. Use [TAB} your arrow keys to hit the OK dialog button and then select ACCEPT.
TAB > ENTER
First read it using the arrows and only if you agree, accept.
DOWN > ENTER
2.9 Wait a bit longer
You might want to continue watching the NIPS talk.
2.10 Yay it’s finished! Time to reboot
You’ll see that it’s finished when the screen looks like this:
2.11 We installed graphics drivers so we need to reboot the computer.
Type sudo reboot into the terminal then hit [ENTER]
Step 3 - Confirm Lambda Stack installation was successful
3.1 Log back into your system
3.2 Open the Terminal application
Note that this time it’s in the top right of the screen; thanks Lambda Stack!
You may need to close out of the “Welcome to Ubuntu” dialog by hitting Next > Next > Done.
Click the terminal application icon
3.3 Check to see if TensorFlow, Torch, and Caffe have GPU support
To confirm that your Lambda Stack installation was successful run the following commands:
$ python
>>> import tensorflow as tf
>>> s = tf.Session()
...
>>> import torch
>>> import caffe
>>> import caffe2
3.4 Check the existence of your GPUs with nvidia-smi
To see the entire output of nvidia-smi you’ll need to expand your terminal. After you’ve done so, just type in nvidia-smi and hit [ENTER].
This will show that you now have a TensorFlow installation with full GPU support installed thanks to Lambda Stack’s TensorFlow GPU installation for Ubuntu 18.04. Congrats!