Neural Network Workstations

Setting Up , MLFlow Server , Activating setup , Cloud storage

Setting workstation up

The $hostname$ of the new machines are

ceg-cnc197009d.coeit.osu.edu
ceg-dnc197129d.coeit.osu.edu
ceg-dnc197144d.coeit.osu.edu

You need to be member of the following group to be able to access these systems: sec_ceg_yilmaz-15_pcvlab.

You can use this hostname to connect directly via SSH (for a terminal/non-graphical session) or via FastX v3 (for a graphical session).

The machine uses your OSU usernames and passwords for login, as with other OSU services. You must use VPN client Pulse to connect to OSU Engineering college and ssh to the Lambda workstation from there.

Pulse VPN client installation and information can be found here
ssh into system using
ssh ceg-cnc197009d.coeit.osu.edu -l username.number
ssh ceg-dnc197129d.coeit.osu.edu -l username.number

Follow the instructions below to set up your account for deep learning.

Instructions for setting up Anaconda are found here and here
Use pip3 to install pytorch:
pip3 install torch torchvision torchaudio
Instructions for setting up the FastX v3 client are found here
Install jupyter
conda install -c anaconda jupyter

MLFlow for Network Tracking

MLflow server is setup on the ceg-cnc197009d.coeit.osu.edu workstation. To connect to this server, you’ll need to establish an SSH tunnel. Below are the steps to get started:

Step 1: Establish SSH Tunnel

Assuming you are using VS Code and are already connected to one of the workstations, open a terminal and run the following command:

ssh -L 5000:127.0.0.1:5000 username.number@ceg-cnc197009d.coeit.osu.edu

This command creates the required SSH tunnel.

Step 2: Initialize MLflow in Your Code

In your code, initialize MLflow with the following line:

      mlflow.set_tracking_uri("http://127.0.0.1:5000")

You can now use MLflow to log your experiments and metrics.

Step 3: Visualize Experiments in Your Browser

To visualize your experiments and logs, open the following address in your web browser:

http://127.0.0.1:5000

Make sure the SSH tunnel from Step 1 is still active while accessing this address.

Logs and Artifacts

All logs and artifacts are saved to the following path on the server:

/research/nfs_yilmaz_15/mlflow/

Logging in and activating the deep learning environment

ssh into the system and use the following command to activate the environment

conda activate deep_learn

Once you are done, deactivate and logout

conda deactivate

Common cloud storage

The workstations access common cloud storage mounted under

/research/nsf_yilmaz_15

Use this space to store unused data. This data can be accessed by all users of the workstation.