Krystian Safjan izikeros

Anti-hype LLM reading list

Goals: Add links that are reasonable and good explanations of how stuff works. No hype and no vendor content if possible. Practical first-hand accounts of models in prod eagerly sought.

Foundational Concepts

Pre-Transformer Models

Compose file to setup and host a Gitea and Drone environment

This environment is dependent off a common network for docker/git plugin and gitea-server, so for this work fine in closed networks we have to create a custom network before run this composefile.

First create your network:

docker network create gitea-network

Useful Pandas Snippets

A personal diary of DataFrame munging over the years.

Data Types and Conversion

Convert Series datatype to numeric (will error if column has non-numeric values)
(h/t @makmanalp)

	from typing import Generic, TypeVar

	import numpy as np
	from pydantic.fields import ModelField

	JSON_ENCODERS = {
	np.ndarray: lambda arr: arr.tolist()
	}

	DType = TypeVar('DType')

	#!/usr/bin/env python3
	'''2021-03-26: Reverse-engineer by searching for the following terms in features*.js:
	- bracketMaintenanceMarginRate
	- cumFastMaintenanceAmount
	- bracketNotionalFloor
	- bracketNotionalCap'''


	# (max) position, maintenance margin, maintenance amount
	maint_lookup_table = [

	#!/usr/bin/env python3
	from __future__ import annotations

	import json


	class CompactJSONEncoder(json.JSONEncoder):
	"""A JSON Encoder that puts small containers on single lines."""

	CONTAINER_TYPES = (list, tuple, dict)

	\documentclass[15pt]{scrartcl}
	\usepackage[a6paper,left=1cm,right=1cm,top=2cm,bottom=1cm,heightrounded]{geometry}
	\usepackage[svgnames]{xcolor}
	\usepackage{pdflscape}
	\usepackage{setspace}
	\usepackage[utf8]{inputenc}
	\usepackage[T1]{fontenc}
	\usepackage{graphicx}
	\usepackage{wallpaper}
	\usepackage[normalem]{ulem}

	"""
	In jupyter notebook simple logging to console
	"""
	import logging
	import sys

	logging.basicConfig(stream=sys.stdout, level=logging.INFO)

	# Test
	logger = logging.getLogger('LOGGER_NAME')

	"""
	以下の論文で提案された改良x-means法の実装

	クラスター数を自動決定するk-meansアルゴリズムの拡張について
	http://www.rd.dnc.ac.jp/~tunenori/doc/xmeans_euc.pdf
	"""

	import numpy as np
	from scipy import stats
	from sklearn.cluster import KMeans