README.md

---
tags:
- llm
- llama
- llama2
- questions
---

<!-- header start -->
<div style="display: flex; flex-direction: column; align-items: center;">
</div>  
<div style="width: 100%;">
    <img src="https://huggingface.co/FPHam/Reverso_Expanded_13b_Q_Generator_GPTQ/resolve/main/reverso-expanded.jpg" alt="Reverso Expanded 0.2" style="width: 80%; min-width: 200px; display: block; margin: auto;">
</div>
<div style="display: flex; flex-direction: column; align-items: center;">
        <p><a href="https://ko-fi.com/Q5Q5MOB4M">Buy my great moustache Ko-fi</a></p>
    </div>
<!-- header end -->

# Reverso Expanded 0.2 (The Crash-files)

Reversing soon


As you can clearly see from the size of the moustache, Reverso Expanded has much wider reach with his questions. 

The original [Reverso](https://huggingface.co/FPHam/Reverso_13b_Q_Generator_GPTQ) was finetuned to ask mostly "What...?" type question, because that's who he is. "What" type of person. 

But his curious brother, Reverso Expanded, asks many more detailed questions (such as Explain..., Write...., Provide..., What if..., Let's say...)

There are pluses and minuses to this. Reverso Expanded has more refined asking capabilities (just try it, you'd be SHOCKED), but if something is just a blabbing prose without much of aim and can resort to "Write a long and boring article about nothing in particular" type of response, which may not be for everyone, especially if you feed it a lot of text about nothing (basically most of the books in Top 10 Romance category). 

Reverso Expanded also happily crashed my computer during training in about 3/4 of the entire process and so the last recoverable checkpoint was not fully what he wanted to be. That's life. 

He still seems to feel fine, even if not at full brain capacity. Would the full finetune be better? We will never know...