LangModelLab is a 6 lesson learning path designed to help people with basic python understanding get a grasp on how LLMs work. You'll be guided through various language models, ranging from a super simple bigram model to a state-of-the-art multi-modal model. In my experience, the best way to learn is from first-principles, here's your chance to learn LLMs starting at 0!
The course is designed as a step-by-step lab experience through the following progression:
- Bigram Model: Statistical foundations of language modeling | DONE
- N-gram Extensions: Higher-order models and their limitations | IN PROGRESS
- Word Embeddings: Continuous vector representations | COMING SOON
- Recurrent Networks: Sequence modeling with RNNs/LSTMs | COMING SOON
- Transformers: Self-attention and modern architectures | COMING SOON
- Multi-modal Models: Integrating text models that are capable of undertsanding images | COMING SOON
- Provide a clear, incremental learning path from basic statistical models to advanced neural architectures
- Offer well-documented implementations of various models
- Bridge theoretical concepts with practical implementations
- Make complex language model concepts accessible to intermediate Python programmers
The project is not yet ready for general use. Once initial lessons are available:
- Clone the repository
- Navigate to
lessons/01_bigram_model/to begin - Follow the README in each lesson folder to get started