Running a model

nengo_loihi.Simulator is how you run models in nengo_loihi. It works much the same as the regular nengo.Simulator (you can read more about that in the Nengo documentation if you are new to Nengo), with a few extra arguments.

Here is a basic example of creating a Nengo model and running it on Loihi (note: you should first follow the instructions for installing and connecting to Loihi).

with nengo.Network() as net:
    stim = nengo.Node(0.5)
    ens = nengo.Ensemble(10, 1)
    nengo.Connection(stim, ens)
    p = nengo.Probe(ens)

with nengo_loihi.Simulator(net) as sim:
    sim.run(1.0)

Simulator arguments

nengo_loihi.Simulator has two extra parameters (beyond those in the standard nengo.Simulator) that you need to know about.

target

nengo_loihi.Simulator can be used to run your model on a physical Loihi board or on a software simulation of the board. This is specified with the target parameter:

with nengo_loihi.Simulator(net, target="loihi") as sim:

There are currently three valid values for target:

• 'loihi': Runs the model on a physical board (must have NxSDK installed and be connected to a board)
• 'sim': Runs the model using a simulation of the board
• 'simreal': Runs the model using a simulation of the board with floating point precision (as opposed to the integer discretization present on the real board)

precompute

You can create your Simulator with

with nengo_loihi.Simulator(net, precompute=True/False) as sim:

• Use precompute=True if your model does not require live interaction between the Loihi board and the host. This will generally simplify and speed up the simulation of your model.
• Use precompute=False if your model does require live interaction (e.g., you have a nengo.Node that is receiving input from the simulation and doing some processing, as in online learning). This will generally be slower.