XythicK/IoraX-3B

🚀 IoraX 3B — Efficient Conversational AI Model

✨ Model Overview

IoraX 3B is a highly efficient 3-billion parameter Transformer, fine-tuned using LoRA adapters on Meta LLaMA 3.2 (3B) — with 4-bit quantization to keep it lightning fast and lightweight!

This model specializes in deep conversational understanding, logical reasoning, and coherent long-form generation — your AI companion for research, education, and creative tasks.

---

🎯 Features & Capabilities

• 🧠 Size: 3B parameters
• ⚙️ Base: Meta LLaMA 3.2 (3B)
• 🔧 Fine-tuning: LoRA with 4-bit quantization
• ⏳ Max context length: 2048 tokens (with RoPE scaling)
• 📚 Training data: Blend of public conversational datasets + expert-curated Q&A
• 🔄 Epochs: 3 for balanced speed and learning
• 🌍 Language: English

---

🚀 Use Cases

Use Case	Description
💬 Conversational AI	Customer support, chatbots, assistants
🎓 Education	Tutoring, concept explanation, Q&A
🧪 Research Assistant	Drafting, summarizing, brainstorming
✍️ Creative Writing	Storytelling, script generation

---

⚠️ Limitations

• 📅 Knowledge cutoff: Data up to 2023 only
• ⚖️ Bias: May reflect biases present in the training corpus
• ✔️ Accuracy: Verify important outputs, especially in critical domains
• 🧑‍⚖️ Not a replacement for experts: Use responsibly

---

💡 Quick Start

from transformers import AutoTokenizer
from unsloth import FastLanguageModel

model_name = "XythicK/IoraX-3B"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = FastLanguageModel.from_pretrained(model_name, load_in_4bit=True, max_seq_length=2048)

messages = [
    {"role": "user", "content": "Explain the philosophical significance of the Eiffel Tower. 🌉🤔"}
]

inputs = tokenizer.apply_chat_template(
    messages, 
    tokenize=True, 
    add_generation_prompt=True, 
    return_tensors="pt"
).to("cuda")

outputs = model.generate(
    input_ids=inputs, 
    max_new_tokens=128, 
    temperature=1.2, 
    use_cache=True
)

print(tokenizer.batch_decode(outputs, skip_special_tokens=True))

🙋 Contact

Maintainer: M Mashhudur Rahim [XythicK]

Role:
Independent Machine Learning Researcher & Model Infrastructure Maintainer

(Focused on model quantization, optimization, and efficient deployment)

For issues, improvement requests, or additional quantization formats, please use the Hugging Face Discussions or Issues tab.

📄 Citation

If you use IoraX in your work, please cite:

@misc{ioraX2025,
  title = {IoraX 3B: Efficient Conversational AI},
  author = {M Mashhudur Rahim (XythicK)},
  year = {2025},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/XythicK/IoraX-3B}}
}

❤️ Acknowledgements

Thanks to Hugging Face and the open-source machine learning community for providing the tools and platforms that make efficient model sharing and deployment possible.