Simple AI Apps using Gradio

Apps developed by the WFU HPC Team
Cody Stevens (stevca9@wfu.edu)
Adam Carlson (carlsoas@wfu.edu)
Sean Anderson (anderss@wfu.edu, primary contact)

Here are two simple AI OOD apps using the Gradio framework. Gradio apps are created with Python scripts that contain functions and classes that define the different elements within the interface. It has excellent documentation with many examples for different interfaces.

Python Environment

You first need to create a new Python virtual environment with the required dependencies. Use any version of Python 3 that known good and available on your HPC system. This example is saved in the $HOME directory, for simplicity, which is not best practice.

python3 -m venv ${HOME}/env-chatbot # create environment
. ${HOME}/env-chatbot/bin/activate  # activate environment

Now, install the required modules into the environment. The specific versions are neeeded by the Audiocraft app, but feel free to experiment as you see fit.

python3 -m pip install -U numpy==1.26.4 \
                          gradio==4.44.1 \
                          pydantic==2.10.6 \
                          llama_cpp_python \
                          openai \
                          audiocraft

Chatbot App: Pre-Trained Model

Hardware requirements: By using a quantized pre-trained model, this chatbot app can run on as little as 8 CPU cores with 8 GB of RAM. It will not be the fastest, but works pretty well!

We also need to download a pre-trained LLM for the chatbot app. For this example, we'll use a pre-processed model from HuggingFace that can be downloaded directly in the GGUF format. It up to you to agree to their license terms as needed.

mkdir -p ${HOME}/llm
wget https://huggingface.co/professorf/Meta-Llama-3-1-8B-Instruct-f16-gguf/resolve/main/llama-3-1-8b-instruct-f16.gguf -O ${HOME}/llm/llama-3-1-8b-instruct-f16.gguf

This model is around 16 GB so plan accordingly!

AudioCraft App: Script and Demos

Harware requirements: We have tested this app using 16 CPU cores, 16 GB of RAM, and one NVIDIA A100 40 GB card. It should work using a variety of GPU offerings.

See the AudioCraft GitHub for the demo script and other demos.