Nengo 2.4.0 released

The Nengo team is stoked to announce the release of Nengo 2.4.0!

What is Nengo?

Nengo is a Python library for building and simulating large-scale neural models for AI and robotics. It can be thought of as a neural compiler, transforming a functional description of a neural model to a network of spiking or non-spiking neurons that can run on multiple backends including GPUs and neuromorphic hardware.

What’s new?

The biggest change in 2.4.0 is an optimization step in the build process. When you do nengo.Simulator(model), Nengo sets up some internal data structures to run the model quickly. Now, after setting up those internal data structures, we inspect them and merge together similar operations in order to speed up the simulation. The end result is that many common types of models will be faster in Nengo 2.4.0, so we recommend upgrading as soon as possible!

Note, however, that optimizing models does take some time, so if you’re running into long build times and you’re not running your model for very long, you can turn off the optimizer like so:

nengo.Simulator(model, optimize=False)

Along with adding the optimizer, we have fixed a few bugs, and now raise exceptions when model act strangely. For example, if a node returns np.nan or np.infinity, we will now raise a SimulationError, as these issues are hard to track down and debug.

To see the full list of changes in Nengo 2.4.0, head to the Github release page.

How do I get it?

To get the new version of Nengo, use pip.

pip install --upgrade nengo

Where can I learn more?

Where can I get help?

You’re already there! If you have an issue upgrading or have any other questions, please post them in this forum.

1 Like