SicariusSicariiStuff's picture
Update README.md
0610f08 verified
---
base_model:
- SicariusSicariiStuff/Negative_LLAMA_70B
- invisietch/L3.1-70Blivion-v0.1-rc1-70B
- EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1
- aaditya/Llama3-OpenBioLLM-70B
library_name: transformers
tags:
- merge
- axolotl
- finetune
license: llama3.3
license_name: llama3.3
language:
- en
---
<html lang="en">
<head>
<meta charset="UTF-8" />
<title>Pernicious Prophecy 70B</title>
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link
href="https://fonts.googleapis.com/css2?family=Darker+Grotesque:wght@300..900&family=Uncial+Antiqua&display=swap"
rel="stylesheet">
<style>
html,
body {
margin: 0;
padding: 0;
background: rgb(11, 15, 25);
color: #E6FFE6;
font-family: 'Darker Grotesque', sans-serif;
}
@keyframes runeGlow {
0% {
text-shadow: 0 0 4px #91ca00;
filter: brightness(0.7);
}
50% {
text-shadow: 0 0 8px #91ca00;
filter: brightness(1.0);
}
100% {
text-shadow: 0 0 4px #91ca00;
filter: brightness(0.7);
}
}
img.badge {
filter: grayscale(100%);
transition: filter 0.7s ease-in-out;
}
img.badge:hover {
filter: grayscale(0%);
}
.rune-border::before,
.rune-border::after,
.vertical-sides::before,
.vertical-sides::after {
animation: runeGlow 1.5s infinite alternate;
}
.rune-border::before {
animation-delay: 0s;
}
.rune-border::after {
animation-delay: 0.2s;
}
.vertical-sides::before {
animation-delay: 0.4s;
}
.vertical-sides::after {
animation-delay: 0.6s;
}
.rune-border {
position: relative;
max-width: 45em;
margin: 2em auto;
padding: 2em 4em;
box-sizing: border-box;
}
.rune-border::before,
.rune-border::after {
position: absolute;
left: 0;
right: 0;
margin: 0 2em;
text-align: center;
white-space: nowrap;
overflow: hidden;
color: #91ca00;
text-shadow: 0 0 4px #91ca00;
font-family: monospace;
font-size: 14px;
content: "ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ";
}
.rune-separator:after {
position: absolute;
left: 0;
right: 0;
margin: 0 2em;
text-align: center;
white-space: nowrap;
overflow: hidden;
color: #91ca00;
text-shadow: 0 0 4px #91ca00;
font-family: monospace;
font-size: 14px;
content: "ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ";
}
.rune-border::before {
top: 0;
}
.rune-border::after {
bottom: 0;
}
.vertical-sides {
position: absolute;
margin: 2em 0;
top: 0;
bottom: 0;
left: 0;
right: 0;
pointer-events: none;
}
.vertical-sides::before,
.vertical-sides::after {
position: absolute;
top: 0;
bottom: 0;
width: 1.5em;
white-space: nowrap;
overflow: hidden;
color: #91ca00;
text-shadow: 0 0 4px #91ca00;
font-family: monospace;
font-size: 14px;
writing-mode: vertical-rl;
text-orientation: mixed;
}
.vertical-sides::before {
left: 0;
content: "ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ";
}
.vertical-sides::after {
right: 0;
content: "ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ | ᛁᛏ ᛁᛋ ᚢᚱᛁᛏᛏᛁᚾ ᛅᚾᛏ ᛁᛏ ᚢᛁᛚᛚ ᚴᚬᛘᛁ ᛏᚬ ᛒᛅᛋᛋ";
}
h1,
h2,
h3 {
font-family: "Uncial Antiqua", serif;
font-weight: 400;
font-style: normal;
color: #426100;
-webkit-text-stroke: 1px #91ca00;
text-stroke: 1px #91ca00;
margin-top: 1em;
}
h2 {
padding-top: 1.5em;
}
a {
color: #619300;
text-decoration: none;
}
a:hover {
text-decoration: underline;
}
h1 {
font-size: 2.5em;
}
h2 {
font-size: 2em;
}
h3 {
font-size: 1.5em;
}
p,
li {
font-size: 1.2em;
line-height: 1.2;
}
p.red {
color: #ef2323;
}
img {
border-radius: 20px;
max-width: 100%;
height: auto;
display: block;
margin: 0 auto;
}
.sidebyside {
display: flex;
justify-content: center;
/* Center horizontally */
align-items: center;
/* Align images vertically */
gap: 1em;
/* Space of 1em between images */
flex-wrap: wrap;
/* Wrap to next line if needed */
}
.sidebyside img {
max-width: 100%;
/* Ensure images are responsive */
height: auto;
/* Maintain aspect ratio */
display: inline;
}
.container {
display: flex;
flex-direction: column;
align-items: center;
justify-content: center;
text-align: center;
}
</style>
</head>
<body>
<div class="rune-border">
<div class="vertical-sides"></div>
<div class="container">
<h1>Pernicious Prophecy 70B</h1>
<p>
<img src="./header.gif" alt="Pernicious Prophecy 70B GIF" />
</p>
<h2 style="margin-top: 0em; padding-top: 0em;">Jump Straight In...</h2>
<p>
<a href="#settings">Click here for downloads & settings</a>
</p>
</div>
<div class="rune-separator"></div>
<h2 style='padding-top:0.5em;'>An Introduction...</h2>
<p>
<b>Pernicious Prophecy 70B</b> is a Llama-3.3 70B-based, two-step model designed by <a
href="https://huggingface.co/Black-Ink-Guild">Black Ink Guild</a> (<a
href="https://huggingface.co/SicariusSicariiStuff">SicariusSicariiStuff</a> and <a
href="https://huggingface.co/invisietch">invisietch</a>) for uncensored roleplay, assistant tasks, and general
usage.
</p>
<p class="red">
<b>NOTE:</b> Pernicious Prophecy 70B is an uncensored model and can produce deranged, offensive, and dangerous
outputs. You are solely responsible for anything that you choose to do with this model.
</p>
<p>
If you have any issues or just want to chat about Pernicious Prophecy &amp; future Black Ink Guild releases, join
<a href="https://discord.gg/gXQzQcnedb">our Discord server</a>.
</p>
<div class="rune-separator"></div>
<h2 id="settings">Engage the Model...</h2>
<h3>Model Downloads</h3>
<p>
FPX:
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B">FP16 (HF)</a> |
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B_FP8">FP8 (Aph.)</a>
</p>
<p>
GGUF:
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B_GGUF_Q4_K_S">Q4_K_S</a> |
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B_GGUF_Q4_K_M">Q4_K_M</a> |
<a href="https://huggingface.co/mradermacher/Pernicious_Prophecy_70B-GGUF">mradermacher</a>
</p>
<p>
EXL2:
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B-3.5bpw">3.5bpw</a> |
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B-5.0bpw">5.0bpw</a>
</p>
<p>
GPTQ:
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B_GPTQ">GPTQ-4Bit-g32</a>
</p>
<h3>Recommended Settings</h3>
<p>
Pernicious Prophecy 70B uses the Llama-3 Instruct format, which is available as a preset in all good UIs. The
sampler settings used in testing are as follows:
</p>
<ul>
<li><b>Instruct Template</b>: Llama-3 Instruct</li>
<li><b>Context</b>: 32,768</li>
<li><b>Temperature</b>: 0.9-1.1</li>
<li><b>Min P</b>: 0.06-0.12</li>
<li><b>Rep Pen</b>: 1.07-1.09</li>
<li><b>Rep Pen Range</b>: 1,536</li>
</ul>
<p>
Feel free to use other sampler settings, these are just sane defaults. XTC is good for roleplaying with the model
but may not be beneficial for other tasks.
</p>
<h3>Context Length</h3>
<p>
The model has been tested in roleplays using up to <b>32,768 token context</b> at various quantizations and is
incredibly stable at this context length.
</p>
<p>
It is possible that the context works at even longer context lengths, but it was not deemed within the parameters
of our testing.
</p>
<div class="rune-separator"></div>
<h2>Sip the Poison...</h2>
<p>
Here, you can find example outputs from the LLM to various instructions. For each of these examples, the model was
inferenced at fp8 with 1.0 temperature, 0.1 min-p, 1.04 repetition penalty, and all other samplers neutralized.
</p>
<ul>
<li>
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B/blob/main/nasa.md">Write a 2000 word, Markdown-formatted, report for NASA. Evaluate each of Jupiter's moons as a suitable
colony with pros & cons, then provide a recommendation.</a>
</li>
<li>
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B/blob/main/tone.md">Write me a 3,000 word opening chapter of a 'gritty hard sci-fi' novel, drawing inspiration from
the writing styles of Isaac Asimov & Andy Weir. Use third person personal. Include dialogue and internal monologues.
The POV character for the opening chapter should be a 26 year old astronaut called Tone on a mission to Europa, who
has just realised that the craft for the return journey is broken beyond repair, and he only has supplies for a few
months. Given that survival is impossible, he seeks to spend the few months he has researching titan, so his life
&amp; mission are not wasted.</a>
</li>
<li>
<a href="https://huggingface.co/Black-Ink-Guild/Pernicious_Prophecy_70B/blob/main/cookie.md">Build me a basic cookie clicker game in HTML & Javascript.</a><br />
</li>
</ul>
<p>
These examples were all the best of 2 responses.
</p>
<div class="rune-separator"></div>
<h2>The Codex...</h2>
<p>
Here, you can find some useful prompting tips for working with Pernicious Prophecy 70B.
</p>
<h3>Formatting</h3>
<p>
'Use markdown' and 'use formatting' are likely to produce the best formatted output. We decided to train these on
trigger words to avoid random Markdown in roleplay replies.
</p>
<h3>System Prompting</h3>
<p>
Pernicious Prophecy 70B is very sensitive to prompting, even over long context. The more you instruct it, the more
it will know what you want it to do.
</p>
<p>
'Avoid purple prose, avoid cliches, avoid deus ex machinae' is a useful prompt snippet for roleplaying purposes.
For best results, don't use your roleplay prompt when using Pernicious Prophecy as an assistant.
</p>
<div class="rune-separator"></div>
<h2>Assembling the Repertoire...</h2>
<p>
We used a two-step process: a merge step to combine the abilities of some of the best L3 70B models on Huggingface
and a gentle SFT training step to heal the merge and address some issues around refusals and positivity bias.
</p>
<h3>The Merge Step</h3>
<p>
First, a
<code>model_stock</code> merge was applied using four high-quality Llama-3 based models:
<ul>
<li>
<b>SicariusSicariiStuff/Negative_LLAMA_70B</b> - chosen to be the base model, because of its low censorship,
reduced positivity bias, and engaging writing style
</li>
<li>
<b>invisietch/L3.1-70Blivion-v0.1-rc1-70B</b> - added for its exceptional formatting, roleplay performance,
and general intelligence.
</li>
<li>
<b>EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1</b> - selected for its ability in longer-form storytelling, varied
outputs, and quality thought.
</li>
<li>
<b>aaditya/Llama3-OpenBioLLM-70B</b> - to add a better understanding of anatomy, and another long-form reasoning
model to the stack.
</li>
</ul>
</p>
<h3>The Finetuning Step</h3>
<p>
We used a <b>qlora-based</b>, targeted finetune on 2x NVIDIA RTX A6000 GPUs, with a curated dataset of
approximately 18 million tokens designed to surgically address issues that we identified in the merge.
</p>
<p>
The finetuning took a total of about 14 hours, using Axolotl, and targeted specific high-priority LORA modules
which allowed us to maintain a 16k sequence length even with 96GB VRAM.
</p>
<div class="sidebyside" style="padding-bottom:2em;">
<a href="https://github.com/arcee-ai/mergekit">
<img
class="badge"
src="https://huggingface.co/Black-Ink-Guild/READMETEST/resolve/main/mergekit.png"
alt="Built with Mergekit"
width="200"
height="32"
/>
</a>
<a href="https://github.com/axolotl-ai-cloud/axolotl">
<img
class="badge"
src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png"
alt="Built with Axolotl"
width="200"
height="32"
/>
</div>
</div>
</body>
</html>