nanochat
nanochat is a Python-based project designed for developing, training, and deploying custom Generative Pre-trained Transformer (GPT) models, with a particular focus on chat applications. It encapsulates the full lifecycle of an LLM, from core model implementation and efficient data processing to various training paradigms (pre-training, supervised fine-tuning, reinforcement learning), comprehensive evaluation, and interactive user interfaces via command-line or web. It primarily operates as a backend worker job system, orchestrating complex machine learning tasks.
- Category
- LLM Framework
- Tech stack
- Python, FastAPI
- Tags
- gpt, ml, python, ai