hf_model_readme.md

metadata

license: apache-2.0
datasets:
  - SouthpawIN/senter-omni-data
language:
  - en
base_model:
  - unsloth/Qwen2.5-Omni-3B-GGUF
tags:
  - any-to-any
pipeline_tag: text-generation

🎭 Senter-Omni

Multimodal AI Assistant with Cross-Modal Embeddings

🌟 Overview

Senter-Omni is a 4B parameter multimodal AI assistant that understands and reasons across text, images, audio, and video simultaneously. Built on Qwen2.5-Omni with extended 128K context and Apache 2.0 licensing.

✨ Key Features

• 🎯 ONE MODEL, ALL MODALITIES - Single model for text, image, audio, and video
• ⚡ TRUE STREAMING - Real-time token generation (~0.234s time-to-first-token)
• 🔓 OPEN & UNCENSORED - Apache 2.0 licensed with unrestricted responses
• 🧠 128K CONTEXT - Extended RoPE scaling for massive documents
• 💾 MEMORY EFFICIENT - 4-bit quantized model for consumer GPUs
• 🔍 CROSS-MODAL EMBEDDINGS - Unified 1024D space for all modalities

🚀 Quick Start

from omni import OmniClient

# Initialize Senter-Omni
client = OmniClient()

# Multimodal chat
response = client.chat([
    {"role": "user", "content": [
        {"type": "image", "image": "photo.jpg"},
        {"type": "text", "text": "What do you see?"}
    ]}
])

# Cross-modal embeddings
embedding = client.embed("any content", modality="auto")

📊 Model Specifications

• Parameters: 4B (quantized to 4-bit)
• Context Length: 128K tokens (RoPE scaled)
• Memory Usage: ~8GB VRAM
• Modalities: Text, Image, Audio, Video
• License: Apache 2.0

🔗 Links

• GitHub Repository: https://github.com/SouthpawIN/senter-omni
• Training Dataset: https://huggingface.co/datasets/SouthpawIN/senter-omni-data
• Demo Script: Run python senter_omni_demo.py in the GitHub repo

🎯 Performance

• Time to First Token: ~0.234s
• Text Generation: 2-5 seconds
• Image Analysis: 3-6 seconds
• Audio Processing: 4-8 seconds
• Multimodal Chat: 5-10 seconds

🛠️ Installation

git clone https://github.com/SouthpawIN/senter-omni.git
cd senter-omni
pip install -r requirements.txt
python senter_omni_demo.py

📝 Citation

@misc{senter-omni,
  title={Senter-Omni: Multimodal AI Assistant with Cross-Modal Embeddings},
  author={Chris at Alignment Lab AI},
  year={2024},
  url={https://github.com/SouthpawIN/senter-omni}
}

---

Built with ❤️ by Chris at Alignment Lab AI