Certain packages should be installed in the system:
sudo apt install gcc g++
Needless to say, some distribution of Python 3 should be installed as well (ideally either Anaconda or Miniconda).
Environment instructions can be found in the script createEnv.sh.
If run directly, this script will create a conda environment HPC and install the necessary packages.
However, it should be possible to install the packages directly, since pip3 is used throughout.
If you do use conda, you will need to activate the environment before running any of the Python scripts:
conda activate HPC
This is a very simple CNN to predict classes on CIFAR10. This will download the CIFAR data to a folder data. On completion, it will also save the final model to models/model.pth. The script should otherwise run in a self-contained fashion.
python simpleCNN.py
On a GeForce GTX 4060, this took less than 2 minutes to train. Final training and testing accuracy should be somewhere in the 70s, maybe 80s.
This loads up Stable Diffusion XL and its refiner and generates some images. Effectively derived from here.
The script will download a bunch of model files to ~/.cache/huggingface/hub directory.
When complete, it will save ten images to the out_images directory in the repo.