The Quantum Transformer Jupyter Notebook is a notebook designed to demonstrate quantum state encoding and the quantum attention mechanism. It integrates classical and quantum computing principles to enhance attention mechanisms.
GPU Acceleration: Utilizes GPU acceleration (Nvidia T4) for enhanced computational performance, particularly beneficial for quantum simulation tasks.
Libraries and Dependencies: qiskit, matplotlib, cuquantum, and cupy-cuda12x
Quantum State Encoding: Implements a function amplitude_encode to encode classical data vectors into quantum states using amplitude encoding, a technique crucial for preparing quantum states that represent classical information.
Swap Test Implementation: Provides a swap_test function, for estimating the similarity between two quantum states. This function is critical in quantum attention mechanisms, analogous to dot product calculations in classical attention systems.
Quantum Measurement: Includes a measure_state function for measuring the quantum state produced by the swap test and interpreting the results to derive attention scores.
Classical Components Integration: Contains classical implementations of Softmax and Weighted Sum functions.
By: Levent Ozbek