Large Language Models

Welcome to Let’s Talk LLMs, your dedicated hub for unraveling the fascinating world of large language models (LLMs) and the vast potentials of artificial intelligence.

I do general usefullness evaluations for LLMs

Large Language Models

Let’s Talk LLMs began as a simple curiosity—a passion project turned into an information gateway. Over time, we have become the vanguard of in-depth explorations into open source large language models, from the most celebrated to the obscure, making sense of how these intricate AI structures can revolutionize our daily lives and business operations.

The Let's Talk LLMs Mission

While the tech world is inundated with buzzwords and fleeting trends, our mission remains steadfast: to demystify LLMs for everyone, from tech enthusiasts to everyday users. We want you to understand not just the “what,” but also the “how” and the “why.”

Join Us

As we embark on this enlightening journey, we invite you to join us. Whether you're a novice, a seasoned tech guru, or simply curious about the world of LLMs, there's something here for everyone.

Let's Talk LLMs

Large language models (LLMs) are deep learning algorithms that can recognize, summarize, translate, predict, and generate content using very large datasets.

As open source largue language models have grown in size, so have their capabilities. Broadly, LLM use cases for text-based content can be divided up in the following manner:

Large Language Models

Let's Look at Some of Our Tests

30B

Wizard Vicuna Uncensored

20B

U-Amethyst

70B

Stellar Bright

13B

Athena version 4

13B

Wizard Vicuna Uncensored

11B

Minstral OmniMix

FAQ'S

Frequently Asked Questions

LLM stands for Large Language Model. It is a type of deep learning algorithm specifically designed for a range of natural language processing (NLP) tasks. While many might recognize ChatGTP by OpenAI as a renowned LLM, it's backed by a major corporation. Our focus, however, is on open-source models.

These expansive language models undergo training on substantial textual datasets from platforms such as Wikipedia, GitHub, and more. Comprising trillions of words, the caliber of these datasets plays a pivotal role in determining the model's efficiency. During the training phase, the LLM indulges in unsupervised learning. In this phase, it autonomously processes the data without explicit directives. Through this, the LLM's AI algorithm grasps word meanings, discerns the relationships between words, and interprets words in varying contexts. For instance, it can deduce if "right" implies "correct" or denotes the opposite of "left."

ChatGTP is a propiertary software that for many cases works great for people. In fact, most peeopel will find their needs are generic enough that ChatGTP works great for them. And hey I love me some ChatGTP. But there are also many guardrails and limitations using a software owned by someone else. And that's why the world of open source LLMs has begun to thrive. But with some many out there, it's really hard to know what might work best for your specific needs and that's why I decided to talk about that very issue.

Yes. Absolutely, and that's what this website is all about. We put the many available LLMs to the test to see just how well they perform with real-world tasks. If you want to run one of these large language models on your local computer, check out Oobabooga text generation web-UI.

 

That's not an easy question to answer. Best is a relative term. What you need an LLM for might not be the same as what someone else needs it for so in the end, my answer has to be, it really depends on what you need it for. But that's why I created all of this, I wanted to look at the various open-source large language models and see how well they perform for the average user with regular tasks.

I use a text generation webui called Oobabooga. Despite the funny name, it's a great way to run the many open-source LLMs on my own computer. If you aren't a super duper techie person like me, then you might want to give it a try because they have an easy-to-use Windows installer that really just takes a few minutes to get you up and running. It's really as easy as 1> Download the Windows installer and unzip the folder on your computer. 2> Run the start_windows.bat 3> Wait probably 10 minutes (maybe 15) for it to run all the crazy stuff it needs for the first-time installation. 4> Click update_windows.bat and let it do its initial update. 5> Run start_windows.bat again, and you are good to go. 

After that initial installation, it's super fast and only takes a few minutes to load up.

I'm a big fan of Tom Jobbins (TheBloke) and get all of my LLMs from him. You can find the thousands of models he has loaded up on his hugging face profile. 

I know. If the LLM kicks back a typo in its response, I leave it in there. I want you to know the exact response the LLM gives me when we give it a prompt, and yes, sometimes that means it gives a typo or a wrong answer.

No. I'm a regular person just like you who has an utter and complete fascination with all things AI and have for years. With the release of ChatGTP, I really saw the future and potential in the many applications but found that sometimes the information can be confusing and overwhelming, so I started sharing the information I found with others. I wanted to know what works for regular people and how it can improve our lives. That's why I created this blog and, later, the YouTube channel. I wanted to see what really was the best for the average person's needs. We're not all computer engineers with degrees in computer science or programming.  I'm just a regular person who has an interest in AI. I want to know what LLM works best for my needs. And I share that information with the world in hopes that it helps others, too.

I mean, how else will you ever know the difference between  Tora, Alma, Zephyr, or the Amethyst LLM?