nengo_spinnaker 0.2.0

Tools for simulating neural models generated using Nengo on the SpiNNaker platform

nengo_spinnaker is a SpiNNaker-based simulator for models built using Nengo. It allows real-time simulation of large-scale models.

Quick Start

Install using pip:

pip install nengo_spinnaker

Settings File

To use SpiNNaker with Nengo you must create a nengo_spinnaker.conf file in either the directory you will be running your code from or, more usefully, a centralised location. The centralised location varies based on your operating system:

• Windows: %userprofile%\nengo\nengo_spinnaker.conf

• Other: ~/.config/nengo/nengo_spinnaker.conf

This file exists to inform nengo_spinnaker of the nature of the SpiNNaker machine you wish to simulate with and how to communicate with it. This file may look like:

### SpiNNaker system configuration
#
# Settings for the SpiNNaker machine which will be used to simulate Nengo
# models.

[spinnaker_machine]
hostname: <host name of the machine here>
width: <width of the machine here>
height: <height of the machine here>

# Required parameters are:
#   - hostname: (string) either the hostname or the IP address of the board
#         containing chip (0, 0).
#   - width: (int) width of the machine (0 <= width < 256)
#   - height: (int) height of the machine (0 <= height < 256)
#
# Optional parameters are:
#   - "hardware_version: (int) Version number of the SpiNNaker boards
#         used in the system (e.g. SpiNN-5 boards would be 5). At the
#         time of writing this value is ignored and can be safely set to
#         the default value of 0.
#   - "led_config": (int) Defines LED pin numbers for the SpiNNaker boards
#         used in the system.  The four least significant bits (3:0) give
#         the number of LEDs. The next four bits give the pin number of the
#         first LED, the next four the pin number of the second LED, and so
#         forth. At the time of writing, all SpiNNaker board versions have
#         their first LED attached to pin 0 and thus the default value of
#         0x00000001 is safe.
#
# For a Spin3 board connected to 192.168.240.253 this section would look
# like:
#
# hostname: 192.168.240.253
# width: 2
# height: 2
# hardware_version: 3
# led_config: 0x00000502
#
# For a Spin5 board connected to 192.168.1.1 this section would look
# like:
#
# hostname: 192.168.1.1
# width: 8
# height: 8
# hardware_version: 5
# led_config: 0x00000001

Using nengo_spinnaker

To use SpiNNaker to simulate your Nengo model first construct the model as normal. Then use nengo_spinnaker.Simulator to simulate your model.:

import nengo_spinnaker

# Build model as normal

sim = nengo_spinnaker.Simulator(network)
sim.run(10.0)

# When done
sim.close()

After running your model you must call close to leave the SpiNNaker machine in a clean state. Alternatively a with block may be used to ensure the simulator is closed after use:

with sim:
    sim.run(10.0)

Some specific configuration options are available for SpiNNaker. To use these:

# Modify config to use SpiNNaker parameters
nengo_spinnaker.add_spinnaker_params(network.config)

Current settings are:

• function_of_time - Mark a Node as being a function of time only.

• function_of_time_period - Provide the period of the Node.

For example:

with model:
    signal = nengo.Node(lambda t: np.sin(t))

nengo_spinnaker.add_params(model.config)
model.config[signal].function_of_time = True

Developers

See DEVELOP.md for information on how to get involved in nengo_spinnaker development and how to install and build the latest copy of nengo_spinnaker.