Need help with netket?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

netket
249 Stars 99 Forks Apache License 2.0 2.4K Commits 38 Opened issues

Description

Machine learning algorithms for many-body quantum systems

Services available

!
?

Need anything else?

Contributors list

logo

NetKet

Release Anaconda-Server Badge Build Status GitHub Issues Paper License Binder Code style: black

NetKet is an open-source project delivering cutting-edge methods for the study of many-body quantum systems with artificial neural networks and machine learning techniques. It is a Python library built on C++ primitives.

Installation and Usage

Netket supports MacOS and Linux. The reccomended way to install it in a non-conda python environment is:

pip install netket[all]
The
[all]
after netket will install all extra dependencies of netket. We reccomend to install netket with all it's extra dependencies, which are documented below. However, if you do not have a working MPI compiler in your PATH this installation will most likely fail because it will attempt to install
mpi4py
, which enables MPI support in netket. If you are only starting to discover netket and won't be running extended simulations, you can forego MPI by installing netket with the command
pip install netket # or pip install netket[jax] 

Netket is also available on conda-forge. To install netket in a conda-environment you can use:

conda install conda-forge::netket
Conda by default ships pre-built binaries for recent versions of python. The default blas library is openblas, but mkl can be enforced. The conda library is linked to anaconda's
mpi4py
, therefore we do not reccomend to use this installation method on computer clusters with a custom MPI distribution.

Extra dependencies

When installing netket with pip, you can pass the following extra variants as square brakets. You can install several of them by separating them with a comma. - '[dev]': installs development-related dependencies such as black, pytest and testing dependencies - '[mpi]': Installs

mpi4py
to enable multi-process parallelism. Requires a working MPI compiler in your path - '[jax]': Installs
jax
to enable jax-based neural networks - '[all]': Installs
mpi
,
jax
and
mpi4jax
which is required to use mpi with jax machines.

Since version 3, in addition to the built-in machines, you can also use Jax and PyTorch to define your custom neural networks.

MPI Support

Depending on the library you use to define your machines, distributed computing through MPI might or might not be supported. Please see below: - netket : distributed computing through MPI support can be enabled by installing the package

mpi4py
through pip or conda. - jax : distributed computing through MPI is supported natively only if you don't use Stochastic Reconfiguration (SR). If you need SR, you must install mpi4jax. Please note that we advise to install mpi4jax with the same tool (conda or pip) with which you installed netket. - pytorch : distributed computing through MPI is enabled if the package
mpi4py
is isntalled. Stochastic Reconfiguration (SR) cannot be used when MPI is enabled.

To check whever MPI support is enabled, check the flags ```python

For standard MPI support

netket.utils.mpi_available True

 For faster MPI support with jax and to enable SR + MPI with Jax machines

netket.utils.mpi4jax_available True

## Major Features

  • Graphs

    • Built-in Graphs
      • Hypercube
      • General Lattice with arbitrary number of atoms per unit cell
    • Custom Graphs
      • Any Graph With Given Adjacency Matrix
      • Any Graph With Given Edges
    • Symmetries
      • Automorphisms: pre-computed in built-in graphs, available through iGraph for custom graphs
  • Quantum Operators

    • Built-in Hamiltonians
      • Transverse-field Ising
      • Heisenberg
      • Bose-Hubbard
    • Custom Operators
      • Any k-local Hamiltonian
      • General k-local Operator defined on Graphs
  • Variational Monte Carlo

    • Stochastic Learning Methods for Ground-State Problems
      • Gradient Descent
      • Stochastic Reconfiguration Method
        • Direct Solver
        • Iterative Solver for Large Number of Parameters
  • Exact Diagonalization

    • Full Solver
    • Lanczos Solver
    • Imaginary-Time Dynamics
  • Supervised Learning

    • Supervised overlap optimization from given data
  • Neural-Network Quantum State Tomography

    • Using arbitrary k-local measurement basis
  • Optimizers

    • Stochastic Gradient Descent
    • AdaMax, AdaDelta, AdaGrad, AMSGrad
    • RMSProp
    • Momentum
  • Machines

    • Restricted Boltzmann Machines
      • Standard
      • For Custom Local Hilbert Spaces
      • With Permutation Symmetry Using Graph Isomorphisms
    • Feed-Forward Networks
      • For Custom Local Hilbert Spaces
      • Fully connected layer
      • Convnet layer for arbitrary underlying graph
      • Any Layer Satisfying Prototypes in AbstractLayer [extending C++ code]
    • Jastrow States
      • Standard
      • With Permutation Symmetry Using Graph Isomorphisms
    • Matrix Product States
      • MPS
      • Periodic MPS
    • Custom Machines
      • Any Machine Satisfying Prototypes in AbstractMachine [extending C++ code]
  • Observables

    • Custom Observables
      • Any k-local Operator
  • Sampling

    • Local Metropolis Moves
      • Local Hilbert Space Sampling
    • Hamiltonian Moves
      • Automatic Moves with Hamiltonian Symmetry
    • Custom Sampling
      • Any k-local Stochastic Operator can be used to do Metropolis Sampling
    • Exact Sampler for small systems
  • Statistics

    • Automatic Estimate of Correlation Times
  • Interface

    • Python Library
    • JSON output

License

Apache License 2.0

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.