The jVMC library allows to easily create Neural Quantum State (NQS) applications. It is targeted at users who wish to be very flexible when creating (new) NQS algorithms, but at the same time don’t want to have the overhead of writing GPU-cluster-ready code. It is based on the JAX library, allowing to conveniently write numpy-like code while don’t falling short on computational performance. The library is described in detail in this paper, while there are detailed examples on GitHub.
- NQS applications based on JAX, reducing the overhead of writing the required functionalities yourself.
- GPU-cluster ready while maintaining readability and convenience of numpy.
- Features ground state searches as well as unitary and dissipative real-time dynamics.