tuxedocat

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.

PythonのPackage Managerを深く知るためのリンク集

以下の発表(2023/10/12)につき作成した、Pythonのパッケージ管理について学ぶ上で有益なリンクを集めたもの。

Pythonでの開発に関するベストプラクティス等を知ることは目的にしていない。
Package Managerを自作したり、開発にコミットするために必要なベースの知識を補うリンク集。

Reinforcement Learning for Language Models

Yoav Goldberg, April 2023.

Why RL?

With the release of the ChatGPT model and followup large language models (LLMs), there was a lot of discussion of the importance of "RLHF training", that is, "reinforcement learning from human feedback". I was puzzled for a while as to why RL (Reinforcement Learning) is better than learning from demonstrations (a.k.a supervised learning) for training language models. Shouldn't learning from demonstrations (or, in language model terminology "instruction fine tuning", learning to immitate human written answers) be sufficient? I came up with a theoretical argument that was somewhat convincing. But I came to realize there is an additional argumment which not only supports the case of RL training, but also requires it, in particular for models like ChatGPT. This additional argument is spelled out in (the first half of) a talk by John Schulman from OpenAI. This post pretty much

GPU Sharing on GKE DaemonSet

NOTE: This is not a Google supported product.

Example Usage

Create a GKE cluster with a GPU node pool:

gcloud container clusters create gpu-sharing-demo --zone us-central1-c

2020/5/31追記: 自分用のメモに書いていたつもりだったのですが、たくさんのスターを頂けてとても嬉しいです。
と同時に、書きかけで中途半端な状態のドキュメントをご覧いただくことになっており、大変心苦しく思っています。

このドキュメントを完成させるために、今後以下のような更新を予定しています。

TODO部分を埋める
書籍を基にした理論・原則パートと、実装例パートを分割
- 現在は4層のレイヤそれぞれごとに原則の確認→実装時の課題リスト→実装例という構成ですが、同じリポジトリへの言及箇所がバラバラになってしまう問題がありました。更新後は、実装時の課題リストを全て洗い出した後にまとめて実装を確認する構成とする予定です。

2021/1/22追記:

	import time
	import os
	import logging
	import random
	from datasets import load_dataset

	class QuantAutoGPTQ:
	def __init__(self, model_name_or_path, output_dir, dataset,
	num_samples=128, trust_remote_code=False, cache_examples=True,
	use_fast=True, use_triton=False, bits=[4], group_size=[128], damp=[0.01],

	import { Client, ServiceError, Metadata, CallOptions, ClientUnaryCall } from '@grpc/grpc-js';
	import { Message } from 'google-protobuf';

	type OriginalCall<T, U> = (request: T, metadata: Metadata, options: Partial<CallOptions>, callback: (error: ServiceError, res: U) => void) => ClientUnaryCall;

	type PromisifiedCall<T, U> = ((request: T, metadata?: Metadata, options?: Partial<CallOptions>) => Promise<U>);

	export type Promisified<C> = { $: C; } & {
	[prop in Exclude<keyof C, keyof Client>]: (C[prop] extends OriginalCall<infer T, infer U> ? PromisifiedCall<T, U> : never);
	}

	[
	{
	"name": "Pok3r macOS (Karabiner fn)",
	"author": "tuxedocat",
	"switchMount": "cherry",
	"switchBrand": "cherry",
	"switchType": "MX3A-L1xx"
	},
	[
	{

	import numpy as np
	import pandas as pd

	#import lightgbm as lgb
	from optuna.integration import lightgbm as lgb

	from sklearn.model_selection import train_test_split
	from sklearn.utils import check_X_y, safe_sqr
	from sklearn.feature_selection.base import SelectorMixin
	from lightgbm import Booster