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
- 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]
- Clean current AMI for saving space [DONE]
- Prepare DGYDLGPUv3
- Prepare another version with remote desktop
- Web access for submitting gpu scripts
- Next check: 2014-12-20
- New build of public image.
- Installation guide check.
- Just now I updated kernel version to
Linux 3.13.0-39-generic
, however, under this kernel, Nvidia's driver cannot be loaded, yourmodprobe
may reportFATAL: Module nvidia not found
(as I'm writing, I installed the latest driver 340.32). Therefore I have to switch down toLinux 3.13.0-37-generic
. You should do following in order to perform this action:-
Reboot your system, press
Shift
key while it's booting up. ChooseAdvanced Options
and select a older version, in my case, I choseLinux 3.13.0-37-generic
. You now should be able to use this old kernel. -
Open a terminal and start to edit
grub
's config.$ sudo /etc/default/grub
-
Change
DEFAULT_GRUB=0
toDEFAULT_GRUB=saved
. And save the modification. -
In you terminal, execute
$ sudo grub-set-default "Ubuntu, with Linux 3.13.0-37-generic"
-
And execute
$ sudo update-grub
-
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-executegrub-set-default
again with correct title and then executeupdate-grub
. -
Now reboot your machine, you should be able to boot your machine from old kernel version.
-
Hu Yuhuang
Advanced Robotic Lab
Department of Artificial Intelligence
Faculty of Computer Science & IT
University of Malaya
Email: [email protected]