JSON Format for Torch Engine Module

A JSON object describing a Torch Engine has following syntax:

{
  "@Skill.Torch-Engine.Module": {
    "id": "Torch Engine",
    "uuid": "848bf8a0-95f6-4d70-a69c-7162ec2f76ef",
    "modelPath": "C:\\Path\\To\\model.pt",
    "deviceType": "CPU",
    "input": {
      ...
    },
    "output": {
      "type": "Float32",
      "sizes": [10],
    },
    "children": [
      ...
    ]
  }
}

Object Members

id	A string value containing the ID of the module. The ID is used to construct the module's primary key.
UUID	A string value containing the unique UUID of the module.
ACL	An optional JSON object containing the element ACL of the module. The format of the block is described here.
modulePath	A string value containing the absolute path of the model.pt .
deviceType	A string value containing the type of the device ("CPU" or "CUDA").
input	A JSON array containing properties of the input for time series or for non time series.
output	A string value containing properties of the output of the model.
type	A string value containing the data type of the model. Float32 Defines input as float 32bit. Float64 Defines input as float 64bit. Int8 Defines input as signed integer 8bit. UInt8Defines input as unsigned integer 8bit. Int16 Defines input as signed integer 16bit. Int32 Defines input as signed integer 32bit. Int6 Defines input as signed integer 64bit. Bool Defines input as boolean value.
sizes	An array which contains integer values for each dimention.
children	A JSON array containing the output

Please remember that each element block requires two layers of {} due to the syntax restrictions of the JSON format.