GGUF investigation (LLM model file format)
GPT-Generated Unified Format (GGUF) used for store vector matrix for most of LLM models (Ollama, llama.cpp, LM studio, GPT4, vLLM and other). This is C++ file format and can not easy reproduce in other languages (therefore for my own LLM model for Javascript I have developed my own file format Create custom LLM in special format (convenient to Javascript and Java) from A1/A2 Deutsch words and training my own custom AI model to books for Deutsch beginners.).
Project to dump LLM vector matrix and investigate steps how LLM applied for AI I have built this project https://github.com/Alex-1367/AI-test/tree/main/InspectAiMatrix. How Vector databases uses LLM models you can see in repository https://github.com/Alex-1367/AI-test/tree/main/InspectAiMatrix, main point is test-rag-verbose.js.
Ai context:
)
|
|