---
language:
- en
license: mit
library_name: quantum-transformer
tags:
- quantum-ai
- transformer
- quantum-inspired
- language-model
metrics:
- perplexity
- entropy-confidence
---
# DEMO
(https://www.youtube.com/watch?v=btVQvQEKVWM/)

# ⚛️ Quantum Transformer (QT-4Q) - The First Pre-trained Quantum AI Model

**Quantum AI QT-4Q** is a groundbreaking, quantum-inspired transformer model designed for efficient language modeling. Unlike classical transformers, this architecture utilizes **quantum gates** (simulated via high-fidelity linear transformations) as the core mechanism for attention and information processing.

This is the **first pre-trained model** released under the Quantum AI project, marking a new milestone in merging Quantum Mechanics with Generative AI.

## Key Features

- **Quantum-Inspired Attention**: Replaces standard dot-product attention with quantum circuit-inspired gates for complex signal processing.
- **State-Space Efficiency**: Operates in a quantum-like state space, allowing for unique information density.
- **Entropy-Based Confidence**: Built-in metrics to calculate model certainty during generation using Shannon Entropy.
- **BPE Tokenization**: Professional Byte-Pair Encoding for superior fluency.
- **KV Caching**: Optimized for high-speed, low-latency inference.

## Architecture Details

| Parameter | Value |
| :--- | :--- |
| **Model Type** | Quantum Transformer (Causal LM) |
| **Qubits** | 4 Qubits (Simulated) |
| **Context Length** | 256 Tokens |
| **Hidden Dimension** | 128 |
| **Layers** | 4 Quantum Blocks |
| **Heads** | 4 |
| **Vocab Size** | 8,192 |

## Getting Started

You can load and use this model directly using the `quantum-transformer` library.

### Installation

```bash
pip install quantum-transformer
```

### Inference

```python
from quantum_transformer.model import QuantumTransformerLM

# Load the first-ever pre-trained Quantum AI model
model = QuantumTransformerLM.from_pretrained("Harishapc01/quantum-qt-4q")

# Generate text with quantum-inspired logic
output = model.generate(
    prompt="Quantum AI is the future of",
    max_new_tokens=100,
    temperature=0.8
)

print(output)
```

## Evaluation & Confidence

One of the unique features of the QT-4Q model is its **Structural Confidence Guard**. During generation, it monitors the entropy of the probability distribution to ensure the output remains coherent and scientifically grounded.

## License

This model is released under the **MIT License**.

## 🤝 Contact & Support

**Quantum AI Team**  
**Developer**: Harisha P C  
**Email**: [reach.harishapc@gmail.com](mailto:reach.harishapc@gmail.com)  
**Website**: [https://harishapc.com](https://www.harishapc.com)  

---
*Developed for the research of Quantum-Inspired Artificial Intelligence.*