L3-TheSpice-8b-v0.8.3 Open-source Large Language Model - Free Deployment for Enjoying Scenario Narrative and Role-playing

L3 TheSpice 8b V0.8.3

Developed by cgato

TheSpice is a large language model based on the Llama3 architecture, specializing in providing unique interactive experiences with particular optimization for scene narration and role-playing functions.

Large Language Model

Transformers

#Narrative Interaction #Character Psychological Analysis #Streamlined Dataset

Downloads 20

Release Time : 4/23/2024

Model Overview

This model integrates multiple datasets, including the Capybara corpus and Claude multi-turn dialogues, undergoing three training cycles to address overfitting issues and fix the tokenizer.

Model Features

Scene Narration

When querying objects or character information within a scene, the model provides narrative descriptions, typically without advancing the plot.

Character Psychological Activity Exploration

Can explore characters' psychological activities and plans, supporting quick retrieval of character summaries.

Seamless Dialogue Continuation

After obtaining scene or character information, dialogue can continue seamlessly.

Small Manually Curated Dataset

Adopts a traditional training approach focused on smaller, manually curated datasets, embodying the 'less is more' philosophy.

Model Capabilities

Text Generation

Scene Narration

Role-playing

Psychological Activity Exploration

Multi-turn Dialogue

Use Cases

Role-playing

Character Psychological Activity Exploration

Users can explore characters' psychological activities and plans to obtain character summaries.

Enhances immersion and depth in role-playing.

Scene Narration

Scene Element Description

Users can observe any element in the scene through specific prompts (e.g., 'What do I see?').

Provides rich scene descriptions, enhancing the narrative experience.

🚀 TheSpice Model

The latest TheSpice model, trained to be more flexible and offer a unique experience.

Now not overtrained and with the tokenizer fix to base llama3. Trained for 3 epochs. The latest TheSpice, dipped in Mama Liz's LimaRP Oil. I've focused on making the model more flexible and provide a more unique experience. I'm still working on cleaning up my dataset, but I've shrunken it down a lot to focus on a "less is more" approach. This is ultimately a return to the form of the way I used to train Thespis, with more of a focus on a small hand - edited dataset.

🚀 Quick Start

The model is ready to use with the specified prompt formats and presets. You can start interacting with it following the guidelines below.

✨ Features

(Examples from 0.1.1 because I'm too lazy to take new screenshots. It's tested though.)

Narration

If you request information on objects or characters in the scene, the model will narrate it to you. Most of the time, without moving the story forward.

You can look at anything mostly as long as you end it with "What do I see?"
You can also request to know what a character is thinking or planning.
You can ask for a quick summary on the character as well.
Before continuing the conversation as normal.

📚 Documentation

Datasets Used

Capybara
Claude Multiround 30k
Augmental
ToxicQA
Yahoo Answers
Airoboros 3.1
LimaRP

Prompt Format: Chat (The default Ooba template and Silly Tavern Template)

image/png

If you're using Ooba in verbose mode as a server, you can check if your console is logging something that looks like this. image/png

{System Prompt}

Username: {Input}
BotName: {Response}
Username: {Input}
BotName: {Response}

Presets

All screenshots above were taken with the below SillyTavern Preset.

Recommended Silly Tavern Preset -> (Temp: 1.25, MinP: 0.1, RepPen: 1.05) This is a roughly equivalent Kobold Horde Preset.
Recommended Kobold Horde Preset -> MinP

📄 License

This project is licensed under the CC - BY - NC - 4.0 license.

⚠️ Disclaimer

Please prompt responsibly and take anything outputted by any Language Model with a huge grain of salt. Thanks!

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご