Storyteller is a project that aims to experiment with how to train an AI model from scratch, fine-tune it, and measure metrics. By building and training a language model, the Storyteller seeks to explore the processes, challenges, and intricacies involved in developing a model that can generate coherent narratives or text continuations based on various literary sources. The project starts by focusing on English language texts, with plans to expand into Turkish literature to diversify language capabilities.
This repository serves as a hands-on guide for those interested in AI model training and development, showcasing how to start with raw data, prepare it, and move through the different stages of training and evaluation. By experimenting with different datasets, including those from the Gutenberg Project, Storyteller aims to create a foundation for language model training while emphasizing good practices and techniques for model optimization.
My motivation depends on the same reason as the Agile Bacı project. I want to learn the Machine Learning and Artificial Intelligence concepts by making my hand dirty. Training a model from scratch is a big deal, I know! However, I'm here to learn all the nuts and bolts: collecting and cleaning data, tweaking model parameters, and pushing the model to see what it can do.
The target is not just creating a fully functional model, I am looking for those "aha!" moments, the insights that make this whole process fun and rewarding.
I am a newbie in this area, so I should need a lot of help in that journey. So far, I learned something from ChatGPT. However, I am determined to learn the calculations and operations behind the scenes.
Apart from this, you can share your ideas and learning material with me by creating an issue. 🤗