apache/mahout

What it solves

Apache Mahout provides an environment for creating scalable and performant machine learning applications. Its current focus, through the Qumat library, is to simplify quantum computing development by providing a unified API for different quantum backends and accelerating the process of encoding classical data into quantum states.

How it works

Qumat acts as an abstraction layer that allows users to build quantum circuits using standard gates (such as Hadamard and CNOT) and execute them across various backends including Qiskit, Cirq, or Amazon Braket without rewriting code. Additionally, it features a Quantum Data Plane (QDP) that uses GPU-accelerated kernels and DLPack for zero-copy tensor transfer, enabling efficient data movement between PyTorch, NumPy, and TensorFlow for quantum machine learning.

Who it’s for

It is designed for developers and researchers working in quantum machine learning and quantum computing who want to write backend-agnostic code and efficiently handle data encoding for quantum states.

Highlights

• Unified Quantum API: Write once and execute quantum circuits on Qiskit, Cirq, or Amazon Braket.
• GPU-Accelerated Encoding: The Quantum Data Plane (QDP) uses CUDA kernels to encode classical data into quantum states.
• Zero-Copy Data Transfer: Integrates with PyTorch, NumPy, and TensorFlow via DLPack to eliminate overhead during tensor transfers.