EDLUT Engine¶

The EDLUT Engine brings the possibility to connect the EDLUT simulator to NRP-core.

The engine implementation is based on the Protobuf over gRPC communication protocol and it accepts two distinct Datapack types which are described below : EdlutDataSpikesDatapack and EdlutDataCurrentsDatapack.

By default there are two datapacks registered for data exchange with an EDLUT engine: “spikes_datapack” and “currents_datapack”. “spikes_datapack” is of type EdlutDataSpikesDatapack and allows users to both inject spikes into the simulated Neural Network and read spikes from it. “currents_datapack” is of type EdlutDataCurrentsDatapack and can only be used for injecting currents into the simulated Neural Network. When “currents_datapack” is requested from the Engine, an empty datapack is returned.

When using the “spikes_datapack” to read spikes, which is done by using it as an input argument of a TF, the datapack contains all spike events that occurred in the EDLUT simulation during the last timestep. The “spikes_time” attribute of the datapack stores an array of spike times, while the “neuron_indexes” attribute contains an array with the indexes of the neurons that produced the spikes.

On the other hand, when the “spikes_datapack” is used to inject spikes into the simulated neural network, typically when it is returned from a TF, the “spikes_time” and “neuron_indexes” attributes must be used to precisely specify which neurons should spike and when the spike events should occur. It is important to note that the “spikes_time” values must be expressed in the global simulation timeframe, rather than being relative to the current simulation time. To achieve the desired behavior, the current simulation time and the length of the Engine timestep must be considered and taken into account when computing the values of the “spikes_time” array.

Similarly, the “currents_datapack” can be used to inject currents into the simulated neural network. In this case, there are also “spikes_time” and “neuron_indexes” attributes in the datapack, which still indicate which neurons should spike and when the event should occur. However, there is an additional attribute,”current_values”, which specifies the amplitude of the spike, behaving the latter as a current.

EDLUT Datapacks¶

This gRPC implementation uses Protobuf DataPacks. Below is shown the protobuf message definition used by the EDLUT Engine:

syntax = "proto3";

package EdlutData;

/*
 * Data coming from edlut datapack
 */
message Spikes
{
    repeated float spikes_time = 5;
    repeated uint32 neuron_indexes = 6;
}

/*
 * Data input to edlut through currents_datapack
 */
message Currents
{
    repeated float spikes_time = 5;
    repeated uint32 neuron_indexes = 6;
    repeated float current_values = 7;
}

Attribute	Description	Python Type	C type
spikes_time	Spikes Firing Time	numpy.array(numpy.float32)	std::array<float>
neuron_indexes	Neuron Indexes Firing	numpy.array(numpy.uint32)	std::array<uint32>

Attribute	Description	Python Type	C type
time	Simulation Time	float	float
spikes_time	Spikes Firing Time	numpy.array(numpy.float32)	std::array<float>
neuron_indexes	Neuron Indexes Firing	numpy.array(numpy.uint32)	std::array<uint32>
current_values	Current Values	numpy.array(numpy.float32)	std::array<float>

Each of these attributes can be accessed under their respective names from the data attribute of both Edlut Datapack types.

Engine Configuration Parameters¶

This Engine type parameters are defined in EDLUT schema (listed here), which in turn is based on EngineBase and EngineGRPC schemas and thus inherits all parameters from them.

To use this engine in an experiment, set EngineType to “edlut_grpc_engine”.

Parameters inherited from EngineBase schema:

Name	Description	Type	Default	Required	Array
EngineName	Name of the engine	string		X
EngineType	Engine type. Used by	string		X
EngineProcCmd	Engine Process Launch command	string
EngineProcStartParams	Engine Process Start Parameters	string	[]		X
EngineEnvParams	Engine Process Environment Parameters	string	[]		X
EngineLaunchCommand		object	{“LaunchType”:”BasicFork”}
EngineTimestep	Engine Timestep in seconds	number	0.01
EngineCommandTimeout	Engine Timeout (in seconds). It tells how long to wait for the completion of the engine runStep. 0 or negative values are interpreted as no timeout	number	0.0

Parameters inherited from the EngineGRPC schema:

Name	Description	Type	Default	Required	Array
ServerAddress	gRPC Server address. Should this address already be in use, simulation initialization will fail	string	localhost:9004

Parameters specific to this engine type:

Name	Description	Type	Default	Required
NetworkFile	Path to the storage of file with the architecture of the network	string		X
WeightsFile	Path to the storage of file containing initial weights of neurons	string		X
NumThreads	Number of threads used for parallel computing	integer	1
SensorialDelay	Sensorial delay needed by EDLUT to compensate delays in communication and compute future spikes	float	0.00
SaveWeightsPeriod	Period of time to save synaptic weights in a file	float	0.00

Schema¶

As explained above, the schema used by the EDLUT engine inherits from EngineBase and EngineGRPC schemas. A complete schema for the configuration of this engine is given below:

{
    "$schema": "http://json-schema.org/draft-07/schema#",
    "title": "Edlut Grpc Engine",
    "description": "Edlut Grpc Engine configuration schema.",
    "$id": "#EdlutGRPCEngine",
    "allOf": [
      { "$ref": "json://nrp-core/engines/engine_comm_protocols.json#/engine_grpc" },
      {
        "properties": {
          "EngineType": { "enum": ["edlut_grpc_engine"] },
          "ProtobufPackages": { "default": ["EdlutData"]},
          "NetworkFile": {
            "type": "string",
            "description": "Path to Edlut Network configuration file"
          },
          "WeightsFile": {
            "type": "string",
            "description": "Path to Neuron Weights file"
          },
          "NumThreads": {
            "type": "integer",
            "default": 1,
            "description": "Number of threads to use if available"
          },
          "SensorialDelay": {
            "type": "number",
            "default": 0.00,
            "description": "Sensorial delay used to compute future spikes"
          },
          "SaveWeightsPeriod": {
            "type": "number",
            "default" : 0.00,
            "description": "Period of time to save synaptic weights in a file"
          }
        },
        "required": ["NetworkFile", "WeightsFile"]
      }
    ]
  }