AWS GPU Instance for Deep Learning

A description for setting up an Amazon EC2 GPU instance for Deep Learning.

Find the description here.

Current information of the image:

Linux Kernel: 3.13.0-40-generic
GPU Driver: 340.65
CUDA Toolkit: 6.5

Updates

Setup guide is uploaded [2014-10-15]
DGYDLGPUv2 is up, you can search this AMI from public images. [2014-10-15]
Added experimental setup script for seting up environment. [2014-11-01]
DGYDLGPUv3 is ready. [2014-12-11]

Todo

Clean current AMI for saving space [DONE]
Prepare DGYDLGPUv3
- Prepare another version with remote desktop
- Web access for submitting gpu scripts

Maintenance Schedule

Next check: 2014-12-20
- New build of public image.
- Installation guide check.

Notes

Just now I updated kernel version to Linux 3.13.0-39-generic, however, under this kernel, Nvidia's driver cannot be loaded, your modprobe may report FATAL: Module nvidia not found (as I'm writing, I installed the latest driver 340.32). Therefore I have to switch down to Linux 3.13.0-37-generic. You should do following in order to perform this action:
1. Reboot your system, press Shift key while it's booting up. Choose Advanced Options and select a older version, in my case, I chose Linux 3.13.0-37-generic. You now should be able to use this old kernel.
2. Open a terminal and start to edit grub's config.
```
$ sudo /etc/default/grub
```
3. Change DEFAULT_GRUB=0 to DEFAULT_GRUB=saved. And save the modification.
4. In you terminal, execute
```
$ sudo grub-set-default "Ubuntu, with Linux 3.13.0-37-generic"
```
5. And execute
```
$ sudo update-grub
```
6. However, this will not actually update your boot kernel becaue "Ubuntu, with Linux 3.13.0-37-generic" is acutally old title. update-grub actually will give you the correct title for the choice. So re-execute grub-set-default again with correct title and then execute update-grub.
7. Now reboot your machine, you should be able to boot your machine from old kernel version.

Contacts

Hu Yuhuang
Advanced Robotic Lab
Department of Artificial Intelligence
Faculty of Computer Science & IT
University of Malaya
Email: duguyue100@gmail.com