ZeroXClem/Llama-3.1-8B-SuperNova-EtherealHermes

Name: Llama-3.1-8B-SuperNova-EtherealHermes
Availability: InStock

🌌 Overview

ZeroXClem/Llama-3.1-8B-SuperNova-EtherealHermes is a cutting-edge fusion of top-tier Llama 3.1 models, meticulously crafted to balance powerful instruction-following, immersive storytelling, and logical reasoning. This merge integrates the strengths of SuperNova, EtherealHermes, and additional high-performance models, resulting in an adaptable and dynamic AI.

This model is governed by the Meta Llama 3.1 Community License Agreement and is optimized for long-form generation, multi-step reasoning, and roleplay applications.

🚀 Key Features

Advanced Instruction Following – Leverages high-context retention for accurate and logical responses.
Enhanced Roleplay & Storytelling – Supports immersive dialogue, lore-building, and dynamic narrative generation.
Long-Form Content Generation – Capable of producing detailed, coherent text over extended passages.
Adaptive Multi-Domain Performance – Handles research, fiction writing, technical content, and conversation seamlessly.
Highly Efficient Processing – Optimized quantization and inference mechanisms ensure smooth deployment.

🧠 Merged Models

This model is the result of a carefully calibrated merge of the following models:

djuna/L3.1-Purosani-2-8B – A high-performance Llama 3.1 model emphasizing instruction-following and contextual coherence.
invisietch/L3.1-EtherealRainbow-v1.0-rc1-8B – Focuses on creative storytelling, world-building, and conversational depth.
ZeroXClem/L3SAO-Mix-SuperHermes-NovaPurosani-8B – A hybrid powerhouse that integrates Hermes3, SuperNova, and Purosani architectures.
ZeroXClem/Llama3.1-Hermes3-SuperNova-8B-L3.1-Purosani-2-8B – Enhances multi-step inference, logical alignment, and long-form composition.

This curated selection ensures the model is equipped with both technical precision and artistic creativity.

🔧 Merge Configuration

The model was merged using Model Stock methodology with bfloat16 precision to ensure a seamless blend of capabilities. The YAML configuration is as follows:

yaml

# Merge configuration for ZeroXClem-Llama-3.1-8B-SuperNova-EtherealHermes using MODEL STOCK

name: ZeroXClem-Llama-3.1-8B-SuperNova-EtherealHermes
base_model: invisietch/L3.1-EtherealRainbow-v1.0-rc1-8B
dtype: bfloat16
merge_method: model_stock
models:
  - model: ZeroXClem/L3SAO-Mix-SuperHermes-NovaPurosani-8B
  - model: ZeroXClem/Llama3.1-Hermes3-SuperNova-8B-L3.1-Purosani-2-8B
  - model: djuna/L3.1-Purosani-2-8B
  - model: ZeroXClem/Llama-3.1-8B-SuperTulu-LexiNova
tokenizer_source: invisietch/L3.1-EtherealRainbow-v1.0-rc1-8B

This ensures logical coherence, creative diversity, and robust performance across various AI tasks.

🛠 How to Use

🔥 Ollama (Quick Inference)

You can run the model using Ollama for direct testing:

ollama run hf.co/ZeroXClem/Llama-3.1-8B-SuperNova-EtherealHermes-Q4_K_M-GGUF

🤗 Hugging Face Transformers (Python)

python

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import torch

model_name = "ZeroXClem/Llama-3.1-8B-SuperNova-EtherealHermes"

# Load tokenizer & model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name, 
    torch_dtype=torch.bfloat16, 
    device_map="auto"
)

# Initialize text generation pipeline
text_generator = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

# Example prompt
prompt = "Describe the significance of AI ethics in modern technology."

# Generate output
outputs = text_generator(
    prompt,
    max_new_tokens=200,
    do_sample=True,
    temperature=0.7,
    top_k=50,
    top_p=0.95
)

print(outputs[0]["generated_text"])

📌 Best Practices

Use System Prompts:
For best performance, add a system instruction before inference:
"Think step by step with logical reasoning before providing any response."
Uncensored Mode:
For more unrestricted output, set the system message to "." or customize it accordingly.
Quantization Considerations:
- Q4 may lead to refusal issues due to loss of fine-tuning alignment.
- F16 or Q8 is recommended for optimal inference quality.

📜 License

This model is released under the Meta Llama 3.1 Community License Agreement.

⚠ Disclaimer: This model is highly compliant and uncensored. It is the user's responsibility to ensure ethical and appropriate usage, especially in public-facing applications.

💡 Future Improvements

Enhanced ethical alignment while preserving model capabilities.
Further fine-tuning for domain-specific reasoning tasks.
Expanded dataset integration for better real-world knowledge representation.

❤️ Special Thanks

A heartfelt thank you to:

djuna for L3.1-Purosani-2-8B.
invisietch for L3.1-EtherealRainbow.
MergeKit Community for advancing open-source merging techniques.
The 🤗 Hugging Face & Open-Source AI ecosystem for continued AI innovation.

Your contributions fuel the progress of next-gen AI models! 🚀💜

📢 Feedback & Contributions

If you encounter any issues, have suggestions, or wish to contribute, feel free to open a discussion or submit a pull request.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	28.41
IFEval (0-Shot)	73.39
BBH (3-Shot)	32.07
MATH Lvl 5 (4-Shot)	17.45
GPQA (0-shot)	5.70
MuSR (0-shot)	11.32
MMLU-PRO (5-shot)	30.50