|
--- |
|
license: apache-2.0 |
|
tags: |
|
- merge |
|
- model_stock |
|
- TIES_merge |
|
- AstralFusion |
|
- TheSpice |
|
- Yggdrasil |
|
- Bluuwhale |
|
- Llama3 |
|
- storytelling |
|
- roleplaying |
|
- instruction-following |
|
- creative-writing |
|
- fantasy |
|
- long-form-generation |
|
- magical-realism |
|
base_model: |
|
- Locutusque/Llama-3-Yggdrasil-2.0-8B |
|
- bluuwhale/L3-SAO-MIX-8B-V1 |
|
- ProdeusUnity/Astral-Fusion-8b-v0.0 |
|
- cgato/L3-TheSpice-8b-v0.8.3 |
|
library_name: transformers |
|
--- |
|
|
|
# Llama-3-Yggdrasil-AstralSpice-8B ๐๐ฟโจ |
|
|
|
**ZeroXClem/Llama-3-Yggdrasil-AstralSpice-8B** is a celestial fusion of **deep storytelling**, **creative writing**, and **instruction-following** capabilities. Merging some of the most powerful models like **Bluuwhale**, **AstralFusion**, **TheSpice**, and **Yggdrasil**, this model has been crafted to deliver versatility, magical realism, and long-form content generation. |
|
|
|
If you're looking to weave intricate narratives, generate insightful text, or engage in immersive roleplay scenarios, this model will help you manifest your wildest creations! |
|
|
|
--- |
|
|
|
### โก **Model Details** |
|
|
|
**Architecture:** Llama 3.0 8B |
|
**Base Model:** Locutusque/Llama-3-Yggdrasil-2.0-8B |
|
|
|
--- |
|
|
|
### โจ **Models Merged:** |
|
|
|
- **[bluuwhale/L3-SAO-MIX-8B-V1](https://huggingface.co/bluuwhale/L3-SAO-MIX-8B-V1)** |
|
*Bluuwhale SAO-MIX* combines RP and storytelling, excelling in immersive content generation. |
|
|
|
- **[ProdeusUnity/Astral-Fusion-8b-v0.0](https://huggingface.co/ProdeusUnity/Astral-Fusion-8b-v0.0)** |
|
A mysterious, cosmic blend that excels in **role-playing** and **long-form content** creation. Perfect for exploring the celestial unknown. |
|
|
|
- **[cgato/L3-TheSpice-8b-v0.8.3](https://huggingface.co/cgato/L3-TheSpice-8b-v0.8.3)** |
|
A model packed with unique narrative twists, infusing storytelling with character depth and realism. It focuses on creative writing, interactions, and exploration. |
|
|
|
- **[Locutusque/Llama-3-Yggdrasil-2.0-8B](https://huggingface.co/Locutusque/Llama-3-Yggdrasil-2.0-8B)** |
|
The backbone of this model, Yggdrasil, brings stability, coherence, and robust instruction-following for even the most complex of prompts. |
|
|
|
--- |
|
|
|
### ๐ **Merge Methodology**: |
|
|
|
This model was merged using the **TIES merge method** to delicately combine the diverse layers of these powerful models. By balancing density and weight distributions, weโve optimized the blend for **instruction following**, **roleplay**, and **narrative creation**. |
|
|
|
```yaml |
|
models: |
|
- model: bluuwhale/L3-SAO-MIX-8B-V1 |
|
parameters: |
|
density: 0.4 |
|
weight: 0.3 |
|
- model: ProdeusUnity/Astral-Fusion-8b-v0.0 |
|
parameters: |
|
density: 0.5 |
|
weight: 0.3 |
|
- model: cgato/L3-TheSpice-8b-v0.8.3 |
|
parameters: |
|
density: 0.5 |
|
weight: 0.2 |
|
- model: Locutusque/Llama-3-Yggdrasil-2.0-8B |
|
parameters: |
|
density: 0.5 |
|
weight: 0.2 |
|
|
|
merge_method: ties |
|
base_model: Locutusque/Llama-3-Yggdrasil-2.0-8B |
|
dtype: bfloat16 |
|
parameters: |
|
normalize: true |
|
out_dtype: float16 |
|
``` |
|
|
|
--- |
|
|
|
### ๐ **Key Features**: |
|
|
|
- **Creative Writing & Roleplaying**: Enhanced with storytelling prowess, ideal for creative prompts and immersive character dialogues. |
|
|
|
- **Instruction Following**: Built upon a solid Llama 3.0 architecture, the model delivers **accurate** and **coherent responses** to a variety of instruction-following tasks. |
|
|
|
- **Narrative Exploration**: Thanks to **AstralFusion** and **TheSpice** models, this model excels in crafting imaginative narratives and detailed scene descriptions, adding layers of depth to text generation. |
|
|
|
- **Magical Realism**: The model blends realism with a sprinkle of magic, taking your content into realms of fantasy and cosmic wonders. |
|
|
|
--- |
|
|
|
### ๐ **How to Use**: |
|
|
|
You can start generating text with this model using the following Python code snippet: |
|
|
|
```python |
|
from transformers import AutoTokenizer, pipeline |
|
import torch |
|
|
|
model = "your-huggingface-repo/Llama-3-Yggdrasil-AstralSpice-8B" |
|
tokenizer = AutoTokenizer.from_pretrained(model) |
|
|
|
text_generator = pipeline( |
|
"text-generation", |
|
model=model, |
|
torch_dtype=torch.bfloat16, |
|
device_map="auto", |
|
) |
|
|
|
prompt = "You are an expert in ancient cosmic magic. Tell me about the legendary AstralSpice." |
|
generated_text = text_generator(prompt, max_new_tokens=512, do_sample=True, temperature=0.9) |
|
|
|
print(generated_text[0]['generated_text']) |
|
``` |
|
|
|
--- |
|
|
|
### ๐ฎ **Use Cases**: |
|
|
|
- **Immersive Roleplay**: From fantasy adventures to deep sci-fi settings, craft characters and dialogues with personality and depth. |
|
- **Storytelling**: Write unique, multi-layered narratives with character-driven interactions, world-building, and plot twists. |
|
- **Instruction-Based Tasks**: Use for education, research, and structured output generation with high precision. |
|
- **Creative Writing**: Generate poems, magical prose, or intricate creative content with seamless storytelling transitions. |
|
|
|
--- |
|
|
|
### ๐ **License**: |
|
|
|
This model is open-sourced under the **Apache-2.0 License**. You are free to use, modify, and distribute it, with appropriate attribution. |
|
|
|
--- |
|
|
|
### ๐ก **Acknowledgments**: |
|
|
|
Special thanks to the creators of the base models: |
|
- **Bluuwhale** for *L3-SAO-MIX-8B-V1*, |
|
- **ProdeusUnity** for *AstralFusion*, |
|
- **Cgato** for *TheSpice*, and |
|
- **Locutusque** for *Yggdrasil*. |
|
Their incredible models made this merge a reality! ๐ฟโจ |
|
``` |