karpathy

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

mlabonne

# Example usage:
# python merge_peft.py --base_model=meta-llama/Llama-2-7b-hf --peft_model=./qlora-out --hub_id=alpaca-qlora

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

import argparse

def get_args():

smartm13

import functools


def st_cache_monitor(func):
    """ A decorator to handle query_cache=hit/miss utility """
    @functools.wraps(func)
    def wrapper_func(*args, _querying_cache=None, **kwargs):
        """ Wrapper to original func to handle special argument _querying_cache """
        if _querying_cache is Ellipsis:
            raise LookupError("_querying_cache=`miss`")