CmdStan
CmdStan is the command line interface to Stan, a package providing
- full Bayesian inference using the No-U-Turn sampler (NUTS), a variant of Hamiltonian Monte Carlo (HMC),
- penalized maximum likelihood estimation (MLE) using optimization, either Newton or quasi-Newton algorithms BFGS and L-BFGS,
- approximate Bayesian inference using automatic differentiation variational inference (ADVI),
- a full first- and higher-order automatic differentiation library based on C++ template overloads, and
- a supporting fully-templated matrix, linear algebra, and probability special function library.
Home Page
Stan's home page, with links to everything you'll need to use Stan is:
Interfaces
There are separate repositories here on GitHub for interfaces:
- CmdStan (command-line/shell interface)
- CmdStanPy (a lightweight interface to CmdStan for Python users)
- CmdStanR (a lightweight interface to CmdStan for R users)
- PyStan (Python interface)
- RStan (R interface)
- MathematicaStan (Mathematica interface)
Source Repository
CmdStan's source-code repository is hosted here on GitHub.
Licensing
The Stan-to-C++ compiler written in OCaml, core Stan C++ code, and CmdStan are licensed under new BSD.
Note that the Stan math library depends on the Intel TBB library which is licensed under the Apache 2.0 license. This dependency implies an additional restriction as compared to the new BSD lincense alone. The Apache 2.0 license is incompatible with GPL-2 licensed code if distributed as a unitary binary. You may refer to the Apache 2.0 evaluation page on the Stan Math wiki.
Installation
- Download the latest release tarball (use the "green" link) from: CmdStan releases
- Unpack the tarball.
- From the folder, type
makefor a quick tutorial on how to build models.
Installation using git
See Getting Started with CmdStan for instructions how to clone both CmdStan and Stan submodule.
Troubleshooting
As of version 2.22, CmdStan has switched to the new Stan-to-C++ compiler, called stanc3. This compiler is intended to be backwards compatible with the existing Stan language and should accept all models that compile under the release 2.21 compiler, (see this list of bug fixes).
If a model that compiled for versions prior to 2.22, please report a bug on the stanc3 repository. Otherwise, report the issue to the CmdStan repository.