Guide

Implement the model

You need to first implement your model with Turing. We will refer to the directory of the model as model_dir. Let’s say there is a file called main.jl, we will refer to this file as the entrance_file of the model:

using Turing

@model function _model(...)
    ...
end;

function model(_input)
    return _model(...)
end

Note:

1. The model is implemented in _model
2. There is a function named model which acturally calls the _model. The model function is a requirements of Coinfer. With a single paramenter _input.

Add Coinfer as a dependency of you project

Project.toml:

[deps]
    ...
    Coinfer = "2494631e-bceb-4539-8821-5341beee86e1"

Run Coinfer.sample

using Coinfer

Coinfer.ServerlessBayes.sample(model_dir)