I put together a quick demo on how to use huggingface models in your own chatbots. In the backend/load_model.py file you can replace the model name in the line model_name = "microsoft/DialoGPT-medium" to any dialogue-capable model to play around with it. The frontend folder has a small React app that allows you to interface with the model.
To set it up, you can either run load_model.py on your console and then start app.py to host the model's API and also the frontend via an npm start after installing the node requirements, or you can easily containerize them as separate containers for frontend and backend via the supplied Dockerfiles. When running with Docker or via localhost, NVIDIA GPU support is enabled (for Docker use the --gpus=all flag to allow this). You can also deploy the dockerized images via Kubernetes with the supplied config.yaml file and expose the service via the supplied service.yaml file.
This is NOT a production-ready system, many npm-related vulnerabilities have not been fixed. It's for learning purposeds only.
Big thanks for huggingface for hosting the models and DialoGPT specifically, and Lucas Bassetti whose React Simple Chatbot NPM-module I've hacked for this to work.
We take a look at what Transfer Learning is, how it works and how you can modify it to teach it something new with Keras!
We perform basic classification on some telco data: Some data cleaning and feature selection methods are demonstrated in addition to multiple models. We also check some methods on how to deal with an unbalanced dataset.
A spiced-up version of SKLearn's "Faces recognition example using eigenfaces and SVMs" tutorial, used for a webinar for MNDWRK. Additional explanation of steps in Hungarian.
Using OpenAI's API to create a multi-language translation tool with placeholder support. Using Gradio for UI, Prometheus for monitoring and Grafana for visualization of said monitoring. All packaged into Docker with optional GPU support for translation evaluation metrics.