low-resource-emotion-recognition

HPC SETUP NOTES

First run the following command on your command line using your student id
ssh <student_id>@login.hpc.dtu.dk

Then load modules by running the following commands
module load python3/3.10.12
module load cuda/12.1

Then check if the modules loaded correctly by running the following command (you should see python and cuda listed)
module list

Then create a directory in your environment
run ls to see available directories and then run cd Desktop to go into desktop
then run mkdir emotion_recognition to create a directory for our project. Run cd emotion_recognition to go into this folder

Then we will create a virtual environment in this directory for the project. For this run the following command
python3 -m venv .venv (make sure you are in the directory of car-seg as it will be easier to have the venv in the same directory as the project)

Now activate the virtual environment by running source .venv/bin/activate (still inside the car-seg directory). You should see (.venv) in front of your command line prompt, this means the virtual environment is active.
You can deactivate by running deactivate and activate back using source .venv/bin/activate again.

Installing packages - example

Then run the following to install pytorch while venv is active
pip3 install torch torchvision torchaudio
python -m pip install -U matplotlib
Any other package can also be installed the same way while the venv is active

Transferring data

Now that we have the environment we will transfer our data onto the server as well
For this open a new terminal on your own local environment and run the following command
You need to replace the first one with the path in your own computer and also change the student id in the second part
scp -r /Users/arime/Desktop/thesis/data sXXXXXX@transfer.gbar.dtu.dk:~/Desktop/emotion_recognition
Now you should wait until everything is copied on

Now that we have the data we can clone the git repo on there
I used the https clone method with a personal token
run git clone https://github.com/Halutz97/low-resource-emotion-recognition.git

Note that it is maybe easier to just copy files one by one manually!

Now we should have the venv, the data and the code setup to run training

WORKING AFTER INITIAL SETUP

First run the following command on your command line using your student id
ssh <student_id>@login.hpc.dtu.dk

Then load modules by running the following commands
module load python3/3.10.12
module load cuda/12.1

Go into the project dir
cd Desktop/emotion_recognition

Run the following to activate venv
source .venv/bin/activate

Start working