dyno.py

Dyno is a Python library for specifying, solving and simulating dynamic macroeconomic models: stochastic DSGE models solved by first-order perturbation, and deterministic perfect-foresight (transition) models solved with a stacked-time Newton solver. Models are written in the human-readable .dyno language (explicit time indices [t], [t-1], [t+1], steady-state values [~]), or imported directly from Dynare .mod files. Dyno checks steady states and Blanchard-Kahn conditions, computes impulse responses, moments and simulations, renders model reports, and comes with Dyno Lab, a JupyterLab extension for editing and solving models interactively.

Documentation: https://econforge.github.io/dyno.py/

Install

Dyno is distributed as a conda package on the econforge channel on prefix.dev. With pixi, in a new or existing project:

pixi init my-project && cd my-project            # skip for an existing project
pixi workspace channel add --prepend https://prefix.dev/econforge
pixi add dyno

--prepend gives the econforge channel priority over conda-forge, so you get the latest Dyno release. For the graphical interface (Dyno Lab), also add JupyterLab and the extension:

pixi add jupyterlab jupyterlab-dyno
pixi run jupyter lab

See the installation guide for optional components (such as the Dynare preprocessor).

Example

Write a model in neo.dyno:

α <- 0.36
β <- 0.99
δ <- 0.025
ρ <- 0.95
γ <- 2.0

z[~] <- 0.0
k[~] <- ((1/β - (1-δ)) / α)**(1 / (α-1))
y[~] <- k[~]^α
i[~] <- δ * k[~]
c[~] <- y[~] - i[~]

z[t] = ρ * z[t-1] + e_z[t]
y[t] = exp(z[t]) * k[t-1]^α
k[t] = (1-δ) * k[t-1] + i[t]
c[t] = y[t] - i[t]
β * (c[t+1]/c[t])^(-γ) * (α * y[t+1]/k[t] + 1 - δ) = 1

e_z[t] <- N(0.01)

Then solve it and compute impulse responses:

from dyno import DynoModel

model = DynoModel("neo.dyno")
model.check()                          # raises if the steady state is wrong
solution = model.solve()               # first-order perturbation
irfs = solution.irfs(type="deviation", T=40)
print(irfs["e_z"][["y", "c", "k", "i"]].head())
solution.plot(type="deviation")        # interactive chart in Jupyter

Dynare files work the same way: DynoModel("model.mod").

Contributing

Contributors work from a clone of this repository. The sections below, and the Develop Dyno page of the documentation, describe the development setup.

Setup

Using pixi (production and development environment)

pixi can be installed on a *nix system using the following command:

curl -fsSL https://pixi.sh/install.sh | sh

For interactive modeling, Dyno features a JupyterLab extension (jupyterlab_dyno / Dyno Lab):

pixi run -e dev jupyter lab

For development (including documentation and unit tests), pixi provides a set of pre-configured tasks.

Model representation explorer (dev tool)

To visualize how the library currently imports and renders every example model (across backends, strict options, and text/HTML/Markdown output), run:

pixi run -e solara explorer

This is a maintainer/contributor tool for tracking the library's progress across its example models — it is not meant for developing or calibrating your own models; use Dyno Lab (above) for that. See the documentation for details.

To build and serve documentation locally with zensical:

pixi run -e dev docs

To run unit and coverage tests:

pixi run test
pixi run cov

Finally, types can be checked with mypy:

pixi run typecheck