owenergy/llama3-sharegpt-10k-voice-ai

Llama 3 8B - ShareGPT 10K Voice AI

This is a LoRA-finetuned version of Meta-Llama-3-8B-Instruct, trained on 10,887 high-quality conversations from the ShareGPT52K dataset.

🎯 Model Overview

• Base Model: meta-llama/Meta-Llama-3-8B-Instruct
• Training Method: LoRA (Low-Rank Adaptation)
• Quantization: 4-bit (bitsandbytes)
• Dataset: RyokoAI/ShareGPT52K
• Training Conversations: 10,887
• Training Steps: 2,043
• Training Epochs: 3
• Adapter Size: ~161 MB

📊 Training Configuration

• Learning Rate: 2e-4
• Batch Size: 4 (effective: 16 with gradient accumulation)
• Gradient Accumulation Steps: 4
• LoRA Rank (r): 16
• LoRA Alpha: 32
• LoRA Dropout: 0.05
• Optimizer: paged_adamw_32bit
• Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
• Trainable Parameters: 41,943,040 (0.52% of total)

🚀 Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Meta-Llama-3-8B-Instruct",
    device_map="auto",
    torch_dtype=torch.float16
)

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "owenergy/llama3-sharegpt-10k-voice-ai")

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Meta-Llama-3-8B-Instruct")

# Generate response
messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Explain quantum computing in simple terms."}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(
    inputs,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.9,
    do_sample=True
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

🎓 Training Data

The model was trained on 10,887 diverse conversations from ShareGPT52K, covering:

• 💻 Technical assistance and programming
• 📚 General knowledge and education
• ✍️ Creative writing and storytelling
• 🔧 Problem-solving and troubleshooting
• 💬 Natural conversation and dialogue
• 🌍 Wide range of topics and domains

✨ What Makes This Model Special

1. High-Quality Data: Trained exclusively on ShareGPT conversations, known for natural and helpful responses
2. Optimized for Voice AI: Designed for conversational applications with natural dialogue flow
3. Efficient LoRA: Only 161MB adapter that works with the base model
4. Well-Balanced: 3 full epochs ensure good learning without overfitting
5. Conversational Excellence: Excels at multi-turn dialogues and context understanding

💪 Model Strengths

• Natural, human-like conversational responses
• Strong context retention across multiple turns
• Helpful and informative explanations
• Creative problem-solving approaches
• Adaptable to various conversation styles

⚠️ Limitations

• Requires base Llama 3 8B model (this is just the LoRA adapter)
• Trained primarily on English conversations
• May inherit biases present in training data
• Requires proper chat template formatting
• Best performance with conversational use cases

📈 Performance

The model shows improved performance over the base model on:

• Conversational coherence
• Response helpfulness
• Natural dialogue flow
• Context understanding
• Multi-turn conversations

🛠️ Use Cases

Perfect for:

• 🎙️ Voice assistants
• 💬 Chatbots
• 📞 Customer service AI
• 🤖 Interactive AI applications
• 📱 Mobile AI assistants
• 🌐 Web-based chat interfaces

📦 Model Files

• adapter_model.safetensors - LoRA adapter weights (~161MB)
• adapter_config.json - Adapter configuration
• tokenizer_config.json - Tokenizer settings
• special_tokens_map.json - Special tokens
• Checkpoints saved every 100 steps

🔧 Hardware Requirements

For Inference:

• GPU: 12GB+ VRAM (with 4-bit quantization)
• RAM: 16GB+ system memory
• Storage: ~5GB (base model) + 161MB (adapter)

Recommended Setup:

• GPU: RTX 3090, RTX 4090, A100, or similar
• With 4-bit quantization: Can run on consumer GPUs

📝 Citation

@misc{llama3-sharegpt-10k-voice-ai,
  author = {owenergy},
  title = {Llama 3 8B ShareGPT 10K Voice AI},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/owenergy/llama3-sharegpt-10k-voice-ai},
  note = {LoRA finetuned on 10,887 ShareGPT conversations}
}

📄 License

This model inherits the Llama 3 license from Meta. Please review the Llama 3 License Agreement.

🙏 Acknowledgments

• Meta for the Llama 3 base model
• RyokoAI for the ShareGPT52K dataset
• HuggingFace for the transformers and PEFT libraries

📧 Contact

For questions, issues, or feedback, please open an issue on HuggingFace.

---

Model Card by: owenergy
Date: December 2025
Status: Production Ready ✅