Undi95/Llama-3-LewdPlay-8B

LewdPlay-8B

Description

This repo contains fp16 files of LewdPlay Llama3, a finetuned model on multiple RP datasets based on meta-llama/Meta-Llama-3-8B-Instruct.

Dataset used

• Undi95/toxic-dpo-v0.1
• NobodyExistsOnTheInternet/ToxicQAFinal
• Aesir [2] & [3 - SFW / 3 - NSFW]
• cgato/SlimOrcaDedupCleaned (reduced)
• Undi95/Capybara-ShareGPT (reduced)
• Airobors (reduced)
• Pippa (cleaned/reduced)
• Bluemoon (cleaned/reduced)
• LimaRP (8k ctx)

Prompt template: Llama3

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{output}<|eot_id|>

SillyTavern files of Llama3 prompt format (if you still don't have them) : Context - Instruct

Usage

Work best with character card well written, with some exemple message, memory, etc...

Support

If you want to support me, you can here.