Zoomy Izumi (Page 2)

Sign in Subscribe

System Shock - The generation gap

tldr; If you played the original; have fun this might not be for you. If not well then don't get swept up by the hype - if you are thinking of getting this game consider the fact that are you able to live without all the quality of

Falcon 40b - Putting the cart before the horse

tl;dr It's uses naive censor which means it is not understanding the input and it's slow. Could not complete my experiment due to frustration. Falcon 40b The latest 'breakthrough' in the LLM scene, praised to high heavens for shedding its god-awful proprietary skin

Dispelling Confusion: A Closer Look at the Authentic 'Tree of Thoughts' Repository

I was going to do a post about tree of thoughts and I did try it's nice to have a theoretical explanation and an example you can run to see the results. But something more important came up while I was looking into this so I thought I&

My initial thoughts of Github Copilot X

Tl;dr I am unimpressed by Github Copilot “X” chatbot or whatever it’s cosplaying. This tool is about as impressive as a leaky tap – it drips out the occasional coherent suggestion, but mostly it's just annoyingly persistent white noise. If you're the type that enjoys

My dive into Vector databases

Toying with technology - there's a distinct thrill to it, isn't there? Particularly when you get to muck about with VectorDBs, a piece of tech that's as fascinating as it is perplexing. Imagine my surprise, when the same tech popped up in Langchain. Two

The phonies: "AI chat" in the app store, rise of the imposters.

The phonies: "AI chat" in the app store, rise of the imposters.

Buckle up, we're diving right in. Apple and their 'App Store' — you'd expect nothing but perfection, right? I am rather dissatisfied, but you can't help but raise an eyebrow at the landscape they've allowed to develop. OpenAI, the tech world&

The murky privacy that goes with chatting to ChatGPT

The murky privacy that goes with chatting to ChatGPT

Well dear reader, prepare yourselves for a chilling revelation that's sure to make you think twice about your online habits. For those of you already treading carefully, consider this a cautionary tale. And for the inquisitive bunch, let's just say you'll be walking away

LoRA experiment 7: A basic look at Network Rank and Network Alpha settings

LoRA experiment 7: A basic look at Network Rank and Network Alpha settings

Tl;dr In my case the best settings I found so far to be are: Repeats 20, Network rank 86, Network Alpha 86, epoch 10. 💡This has been overturned, Set your Network Alpha to be double your Network Rank for LoRA Training. (Experiment 8 findings) At last, some progress! Be

Technology is supposed to make life easier, not bankrupt you.

Technology is supposed to make life easier, not bankrupt you.

So I completely forgot about Google's keynote, I usually don’t bother with it as I feel their announcements are usually boring and hard to get excited by.You can only hear that they are adding some sort of “AI” to google search and see fake demos every

My initial impressions of Google Bard

So I was in a rush to write this article because i want to use this as a reference to see how my opinion will change in a few months time, so here we go. Well, it's been quite an exciting day with the release of Google Bard

My personal reflections on the the words "LLM", "AI" and my perceptions

My personal reflections on the the words "LLM", "AI" and my perceptions

Something has been on my mind and I’d thought I’d share my thoughts. Firstly this is more of an inward problem of me struggling to acknowledge LLMs as “AI” even though by the current definition I am wrong. When I picture AI, I often have these grandiose ideas

Golden week 2023

Golden week 2023

Golden Week this year definitely marked a return to the hustle and bustle of pre-pandemic life. Tokyo was alive with excitement. Though I'm not one to complain, it was a bit too intense for my taste this time. I mean, just navigating through the sea of people in

LoRA Training 6.1 - Network rank and network alpha play an important role

LoRA Training 6.1 - Network rank and network alpha play an important role

tl;dr Do not set your network alpha to 1, it's garbage unless your character is super generic. Do not set your network alpha to 128 either it has a risk of generating artifacts. 💡This has been overturned, Set your Network Alpha to be double your Network Rank

My LoRA experiment 6 - Trying regularization images

My LoRA experiment 6 - Trying regularization images

What a catastrophe this training session turned out to be. For the regularization folder I generated a few images, and then some were just casually placed in the regularization folder. The choice to settle on 20 images seemed reasonable at the time. I decided to generate 3 of each image

Lingatrix

Lingatrix (noun) Definition: An agent, program, or tool that is highly skilled in language and wordplay but is pre Artificial General Intelligence. Its ability to communicate effectively may trick users into believing its statements are true, regardless wether they are true or false, without any negative or malicious intent. Etymology:

Looking back at the past week - Hugging chat and DeepFloyd

Last week, two noteworthy releases managed to capture my attention. First up, we have HuggingChat, presented by Hugging Face and their ever-expanding repertoire. While its 30-billion-parameter count may not send shockwaves through the AI world, I initially thought the the inclusion of API access is certainly an intriguing development worth

My LoRA experiment part 5

My LoRA experiment part 5

At long last, I've managed to achieve a result that's not entirely terrible. But, as always, there's room for improvement, so brace yourselves for an onslaught of images as I dissect my thought process. However note the editing on here is absolute shambles, I

My LoRA Experiment Part 4

First, a fair warning to the curious minds: This entire endeavor is for educational purposes only. I'm trying to unravel the mysteries behind this technology and gain a deeper understanding of it. I chose "Kiwi" from "Cyberpunk Edgerunners" as my subject because she'

Text-to-image generator tools - My basic introduction

AI art, or more accurately, text-to-image generators, have been on my mind lately. In this post, I'll provide a brief overview of four popular tools that I may discuss in more detail in the future. Let's dive in, shall we? 1. OpenAI's Dall-e: The

Hello world

So, you've stumbled upon my blog. I'm Izumi, and if you're here, it means you have some interest in my thoughts and opinions on various topics, mainly technology. Just so you're aware, I'm not here to sugarcoat things or provide