CoolFace
Modelpublic

Black-Ink-Guild/Pernicious_Prophecy_70B

sourceHugging Facellama3.3updated 2y agoView on Hugging Face
23likes49downloads
Model Card

<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/PerniciousProphecy70B">FP16 (HF)</a> | <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70BFP8">FP8 (Aph.)</a> </p> <p> GGUF: <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70BGGUFQ4KS">Q4KS</a> | <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70BGGUFQ4KM">Q4KM</a> | <a href="https://huggingface.co/mradermacher/PerniciousProphecy70B-GGUF">mradermacher</a> </p> <p> EXL2: <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70B-3.5bpw">3.5bpw</a> | <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70B-5.0bpw">5.0bpw</a> </p> <p> GPTQ: <a href="https://huggingface.co/Black-Ink-Guild/PerniciousProphecy70BGPTQ">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/PerniciousProphecy70B/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/PerniciousProphecy70B/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/PerniciousProphecy70B/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>modelstock</code> merge was applied using four high-quality Llama-3 based models: <ul> <li> <b>SicariusSicariiStuff/NegativeLLAMA_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>