Just let me know the need and I can do it for you remotely.

Q: Manully install .deb in terminal
A: sudo dpkg -i package_file.deb

Q: Manully uninstall .deb in terminal
A: sudo apt-get remove package_name

/home is a auto.fs system which I don't fully understand yet.

All home folders are mounted to be a Network File System (NFS), so everything in /home/* is not saved locally but to a NFS somewhere. This LDAP approach makes it possible to login to any Linux desktop and get access to all the same “home” files.

The issue for us is this limits reading/writing speed to the home foler by the newtwork connection speed. Therefore, some installations of software can be very slow, for example the default installation of conda.

Workaround:

• Create a local folder using sudo mkdir /home_local
• Find the user name and group name using id username
• Change the ownership to the local user using chown -R username:groupname /home_local.
• Reinstall stuffs into this /home_local/ directory.

Q: How to determine whether a directory is on an NFS mounted drive?
A: df <directory> https://superuser.com/a/422064

sudo cp -a /source_dir/. /target_dir/ (source)

There are multiple versions. The following one just works fine.

sudo apt install nvidia-driver-535

Other versions like *-server or *-open or *-server-open have subtle differences that are not too much relevant to our research.

After this, in Matlab, call gpuDevice should detect the GPU correctly. This is a helpful link to test and measure GPU performance in Matlab: Measure GPU Performance

Regarding cuda and cudnn, we don't need to install them following NVIDIA's instruction. Since we are mostly using python, it is much easier to handle them using conda.

This requires sudo, so I will install it for everyone.

Check MATLAB Related for more MATLAB tricks (feel free to edit and contribute your MATLAB skills.)

General Python configurations or tricks:

• Python Programming Related

Specific packages:

• Tensorflow Related

GitKraken is installed as a snap package: https://www.gitkraken.com/git-client/try-free

sudo snap install gitkraken --classic

Installation instructions are removed because it requires sudo.

1. If not using ERAUNet or an ethernet cable, install and connect to the campus VPN first.
2. (IMPORTANT) Following this link to speed up the connection.
   • Basically, you need to make sure LAN connection is selected on your RDP client's Experience tab.
3. Using Win10's Remote Desktop Connection to log in.
   • Use the IP of your desktop. (Keep your IP confident. Don't expose/release/share your IP to/with others.)
   • Use your ERAU ID and passwork to log in.

Known issues:

• Color profile autherization pops up sometime or ever time. It's safe to just ignore it. (Should have been fixed)