Go To AIP Data Types

1 Definition 2 Functional Requirements 3 Syntax 4 Semantics 5    Conformance Testing 6     Performance Assessment

1 Definition

A Data Type including the data constituting a Machine Learning Model, which results from the application of training data to a process, and, optionally, an ML Model Qualifier (TFA-MMQ) providing information on Sub-Types, Formats and Attributes of the ML Model.

2 Functional Requirements

An ML Model Object may include:

  1. The ID of the M-Instance to which it belongs.
  2. The ID of the ML Model Object.
  3. The time of the ML Model Object.
  4. The data of the ML Model, as one or more of:
    1. the data themselves;
    2. the URI at which the data are, with their length in bytes;
    3. the ID of the data.
  5. The ML Model Qualifier, specified by MPAI-TFA, providing information on:
    1. Sub-Types: the type of the ML Model, e.g., Neural-Network, and of the neural network, e.g., Transformer.
    2. Formats: the file extension, the framework and the exchange format of the ML Model.
    3. Attributes: the assets, the ML Model Manifest, the regulations and certification, and the validity of the ML Model.
  6. Data Exchange Metadata providing additional information on the ML Model Object.

3 Syntax

https://schemas.mpai.community/AIP/V1.0/data/MLModelObject.json

4 Semantics

Label Description
Header ML Model Object Header – Standard “AIP-MLO-Vx.y”
MInstanceID Identifier of M-Instance.
MLModelObjectID Identifier of the ML Model Object.
MLModelObjectTime Time of the ML Model Object.
MLModelData The set of data constituting the ML Model. Each item is one of the following:
– DataLength Number of bytes of the ML Model data at DataURI (with DataURI).
– DataURI URI at which the ML Model data are (with DataLength).
– Data The ML Model data.
– DataID Identifier of the ML Model data.
MLModelQualifier ML Model Qualifier (TFA-MMQ): Sub-Types, Formats and Attributes of the ML Model.
DataXMData Information about this ML Model Object instance.
DescrMetadata Descriptive Metadata.

5 Conformance Testing

A Data instance Conforms with ML Model Object (AIP-MLO) if:

  1. Its JSON Object validates against its JSON Schema.
  2. Any included JSON Object validates against its JSON Schema.
  3. All Data in the JSON Object:
    1. Have the specified Data Types.
    2. Conform with the Qualifiers signaled in their JSON Schemas.

6 Performance Assessment

Go To AIP Data Types