Appendix. Settings and an example YAML file containing the configuration of the AltAI-1 neuromorphic processing unit

Configuration file settings must be defined to run the neural network on an AltAI ANN2SNN backend.

The configuration file specifies the number of VLSIs, cores, neurons, and synapses in the AltAI-1 neuromorphic processing unit. The configuration file contains the following settings:

• chip includes a cores setting. This cores value sets the number of cores in one VLSI.
• core is a group of the following settings that describe the core:
  • neurons is the number of neurons in the core.
  • axons is the number of input lines.
  • synapses is the number of synapses in the core.
• row includes a chips setting. This chips value sets the number of VLSIs in one module of the AltAI-1 neuromorphic processing unit backplane.
• col includes a chips setting. This chips value sets the number of modules in the AltAI-1 neuromorphic processing unit backplane.
• not_working_chips is an array with the numbers of VLSIs that will not participate in the execution of the neural network on the neuromorphic processing unit—for example, if these VLSIs are missing on the backplane modules.

  This setting must be configured by a qualified Customer engineer, a Kaspersky employee, or a certified integration partner.

Below is an example YAML file containing the default configuration of the AltAI-1 neuromorphic processing unit. In the example below, the neuromorphic processing unit backplane consists of four modules with eight VLSIs in each. Each VLSI contains 16 cores, each of these with 512 neurons, synapses, and input lines.

hw_config.yaml

chip:

  cores: 16

core:

  neurons: 512

  axons: 512

  synapses: 512

row:

  chips: 8

col:

  chips: 4

not_working_chips: []

Page top