本地 LLM 写俳句(1/4)

这是四篇系列文章中的第一篇。我会介绍如何下载、安装和配置软件,让本地运行的大语言模型写俳句……

最近我添置了两台配置相同的 HP Z440 工作站,配备 M4000 GPU。经过研究,我发现这些硬件可以运行一个小型本地 LLM。此前我只是在浏览器或 VS Code 编辑器中,作为终端用户与语言模型交互。

于是我从 Hugging Face 下载了 Qwen3.5 4B 模型,开始研究如何用 llama.cpp 运行它。


构建模型

要让一切跑起来,需要准备几个部分。

我从 Hugging Face 下载 Qwen 模型和 llama.cpp 源代码,构建中间模型,最后将它转换成 llama.cpp 使用的 GGUF 格式。过程虽然涉及一些技术步骤,但并不复杂。最费时间的是下载 18 GB 数据。

然后,我启动 llama_server,在浏览器里打开它……

它就在那里。

一个带聊天框的网页。

我给自己搭了一个聊天机器人。

这件事既酷,又没那么有意思。

毕竟我已经有 ChatGPT,它的知识和能力要强得多。


与它交谈

当然,我首先开始问它各种问题。

我问了一些常识问题。它了解历史人物(例如 Rudolf Steiner)、地理、科学,还有编程……

答案是:它懂得相当多。

光是这一点就已经很令人惊讶。它没有查询 Google,也没有调用网上的巨型 AI 服务。模型和它所知道的东西,就在我自己的电脑里。

然后,我问它是否知道 Patrick Loeber 的 AI 贪吃蛇教程。

它知道我在说什么。

这引起了我的注意。

Patrick Loeber 的项目对我来说很熟悉。我自己的 Snake Lab 项目,正是从研究让 AI 学会玩贪吃蛇的实验发展而来的。所以,我不再只是问模型一些冷知识,而是开始和它聊我真正了解的主题。

它居然能给出有见地的回应。

真酷。


接下来呢?

问题是,我手里仍然只是一个聊天机器人。

我可以提问,它可以回答。这很了不起,但并不是我真正感兴趣的事情。

我想在程序代码里使用 LLM。

于是我写了一个很粗糙的对话循环。

当时并没有什么宏伟计划。我只是想写一个程序:向模型发送提示词,获取回复,然后用这个回复做点事情。

这就引出了一个重要的工程问题:

我应该让它做什么?

显然,我的答案是:

让它为数字写俳句。


随机数字俳句

程序生成一个随机数,再连同“为这个数字写一首俳句”的提示一起发送给 LLM。

模型写出俳句。

程序将结果写入日志文件。

A new age starts today
Apollo touches the moon's face
New world begins to shine

1970

接着,它又生成一个数字。

再一个。

再一个。

很快,我就拥有了一台本地运行大语言模型的电脑,模型在那里沉思随机整数,并为它们写诗。

说不上有什么实际用途。

但它能用。

真正重要的并不是俳句,而是这个循环:

  1. 我的程序生成输入。
  2. 程序构造提示词。
  3. 程序把提示词发送给 LLM。
  4. LLM 生成回复。
  5. 程序捕获回复。
  6. 重复这个过程。

这是我第一次不再只是与 AI 聊天。

而是把 AI 放进了程序内部。


结论

我的第一个本地 LLM 实验没有产生任何实际价值。

它只是在为随机数字写俳句。

但那个小实验回答了我真正关心的问题。

我可以在本地运行 LLM,从自己的软件与它通信,把信息交给它,获取回复,记录回复,再重复整个过程,而无须一直坐在聊天窗口前。

那时候,我还没有 Ax3l。

甚至连完整的构想都还没有。

我只有一个循环。

以及一个写满俳句的日志文件。


链接

2026

Git Automation

Some of you may have read my Git Battles post, where I talked about how I developed and a documented a Git workflow.

My blog in Simplified Chinese

I recently built MyCount to track website visitors to my projects. I found that that over 40% of my visitors were from China, and that 35% of those used a Ch...

MyCount: Website Analytics & Marketing

I have a future project in mind which will require a Dynamic DNS service. I created BMDynIP for that purpose. After building the service, I wanted a simple a...

Animating Ax3l Playing Snake

I’ve been tinkering with various incarnations of Patrick Loeber’s Train an AI to Play Snake tutorial for a while now. It’s my relaxing goto project that’s fu...

Dynamic DNS Service

This project started as a cron job and a flatfile and -honestly- that would work perfectly well for a small home setup. Say, two or three machines. I’m curre...

Local GeoIP Database

The Bear & Moose GeoIP Project provides an easy-to-use GeoIP service for your applications and for interactive lookups.

An LLM to organize my Movies

Our first TV was a small black-and-white thing where, on rare occasions, we’d watch Animal Kingdom. Then Beta and VHS came along, and you could watch them at...

From Prototype to Product (4/4)

Fr3d worked: It could talk to the Snake Lab, propose changes to the RNN configuration, run experiments, look at the results, and try again. It proved the ide...

Prompt Engineering 101 (3/4)

The Snake Lab had become a proper server, with a ZeroMQ API, configuration validation, a database, and a way to run experiments independently of the client. ...

The Evolution of Snake Lab (2/4)

In my last post, I setup a locally hosted LLM running on an old HP Z440 and wrote a crude little conversation loop. I generated random numbers, fed them to t...

Local LLM Writes Haikus (1/4)

This is part one of a four part blog series. In this post, I download, install and setup some software to prompt a local LLM to write haikus…

From Seed to Signal: AI FX

In the first post, AI FX was still mostly a seed: a repo, a PyPI package, and an idea about using automated FOREX tooling as a learning lab. Since then, the ...

Simulating the Power Grid

A potential client pointed me to Empowering the Energy Transition Through Fast and Flexible Network Simulation. The article describes an Open Source project,...

Delphi in a Day

I have a prospective client who runs a small group of technology businesses, at least one of which integrates with clients using the Delphi programming lan...

A Small Experiment: AI FX

I’ve spent some time around trading systemsL At Merrill Lynch, I supported the Order Audit Trail System (OATS), a compliance system required by NASDAQ. Its p...

Back to top ↑

2025

My Art on Etsy

I setup an Etsy Storefront to sell high quality, museum grade prints of some of my artwork.

Python systemctl Wrapper

I’ve been working on Db4E, my Monero deployment, operations and analytics platform. It uses Linux systemd to spin up manage Monero, P2Pool and XMRig instance...

AI Snake Lab

AI Snake Lab is an interactive reinforcement learning sandbox for experimenting with AI agents in a classic Snake Game environment — featuring a live Textual...

The Power of Pictures

You know that saying, “A picture is worth a thousand words.”? In programming, I’d argue it can be worth a thousand lines of code…

Caring About Constants

This post is about using constants in your code. Why it matters, what the benefits are and how you can do it well. I’m sad to say, I learned this the hard wa...

Why I Built Db4E

When I first started experimenting with Monero mining and self-hosted infrastructure, I did what everyone else does: I downloaded the Monero software, the P2...

Releasing with Git

Let me be honest: Git has punched me in the face more times than I can count. Every time I think I’ve got a smooth release flow down, Git yells, “Not so fast...

Documentation Matters

I was burned by opaque documentation: I’ve been building a modern terminal UI in Python using the excellent Textual framework. It’s sharp, fast, and honestly...

Textual Replaces ApexCharts

The original reporting system was a set of static dashboards built using GitHub Pages, JavaScript, and ApexCharts. It pulled data from my mining setup — via ...

Pivot to Textual

I’ve been deep in the trenches with my Db4E project — pulling together Monero mining infrastructure, MongoDB integration, and a bunch of orchestration magic....

Released Db4E SystemD

I recently published a small Python module to PyPI and GitHub: db4e-systemd, a lightweight wrapper around systemctl that lets you control services (start, st...

NIKU Farms

I received my order from NIKU Farms, YAY! It’s a meat delivery that my sister, Corinna Ghaznavi, gifted me this Christmas. Every two months I receive this am...

The Main Website

As part of my goal to be more visible in the world, I setup this domain i.e. https://osoyalce.com. I use it to publish information about the projects I’m wor...

Reviving the Blog

It’s been a while since I started and then abandoned this blog. My last post was in June of 2024!!

Changing the Blog Site’s Layout

I decided that I didn’t like the way that the main blog page looks. I am using GitHub to host the website and am using a fork of the jekyll/minima theme. My ...

Back to top ↑

2024

Aquaponics Update #1

It’s in the construction phase. I need to complete the water circulation process. I bought a fiber glass kit for this (and also to fix my kitchen sink!).

What I’m working on

The XMR Mining Farm complete with this eye-popping chart The Bear and Moose Knowledge Base My resume MyZODB - A Place to Store Everything

Back to top ↑