dynamiqs/dynamiqs

What it solves

Dynamiqs provides a high-performance framework for simulating quantum systems, addressing the need for faster simulations of large-scale quantum systems and simplified gradient-based parameter estimation and quantum optimal control.

How it works

Built with JAX and based on Diffrax, the library implements GPU-accelerated and differentiable solvers for various quantum equations, including the Schrödinger equation, the Lindblad master equation, and the stochastic master equation. It allows users to batch simulations over Hamiltonians, initial states, or jump operators and compute machine-precision gradients of system parameters.

Who it’s for

Physicists, theorists, experimentalists, and machine learning practitioners working on the simulation, calibration, and control of quantum technologies, particularly those dealing with large-scale problems.

Highlights

• GPU Acceleration: High-performance simulations on both CPUs and GPUs.
• Differentiability: Ability to compute gradients and higher-order derivatives (e.g., Hessian) with respect to system parameters.
• JAX Ecosystem: Full compatibility with JAX and a QuTiP-like API for ease of use.
• Advanced Solvers: Support for modern ODE methods, quantum-tailored methods that preserve physicality, and time-varying problems.
• Concurrent Execution: Batching capabilities to run many simulations simultaneously.
• Customization: Support for custom save functions and the ability to implement custom solvers by subclassing.