Skip to content

Latest commit

 

History

History
229 lines (152 loc) · 5.04 KB

spec-v1alpha1.md

File metadata and controls

229 lines (152 loc) · 5.04 KB
Raw

Table of Contents generated with DocToc

OCI Model Configuration Specification

An OCI Model is a Machine Learning/Deep Learning model. This specification outlines the JSON format describing models for use and execution tool and its relationship to filesystem changesets.

This section defines the application/vnd.caicloud.model.config.v1alpha1+json media type.

Terminology

This specification uses the following terms:

Tensor

Tensors are multi-dimensional arrays with a uniform type.

  • name string, OPTIONAL

    Name of the tensor.

  • description string, OPTIONAL

    Description of the tensor.

  • dType string, OPTIONAL

    DType of the tensor.

  • size object, OPTIONAL

    Size of the tensor.

  • opType string, OPTIONAL

    OpType of the tensor, It is only used for PMML.

  • values object, OPTIONAL

    Values of the tensor, It is only used for PMML.

GitRepo

GitRepo is the Git repository.

  • repository string, OPTIONAL

    Git repository URL

  • revision string, OPTIONAL

    Revision of the repository

  • description string, OPTIONAL

    Description of the training script

Properties

Note: Any OPTIONAL field MAY also be set to null, which is equivalent to being absent.

  • created string, OPTIONAL

    A combined date and time at which the image was created, formatted as defined by RFC 3339, section 5.6.

  • author string, OPTIONAL

    Gives the name and/or email address of the person or entity which created and is responsible for maintaining the image.

  • description string, OPTIONAL

    Description of the model

  • tags object, OPTIONAL

    Tags of the model

  • labels object, OPTIONAL

    The field contains arbitrary metadata for the container. This property MUST use the annotation rules.

  • framework string, OPTIONAL

    The framework of the model (e.g. TensorFlow)

  • format string, REQUIRED

    The format of the model (SavedModel, ONNX and so on)

  • size int, REQUIRED

    The size of the model

  • metrics object, OPTIONAL

    The metrics of the model

    • name string, REQUIRED

      Name of the metric

    • value string, REQUIRED

      Value of the metric

  • hyperParameters object, OPTIONAL

    HyperParameters of the model

    • name string, REQUIRED

      Name of the hyperparameter

    • value string, REQUIRED

      Value of the hyperparameter

  • signature object, OPTIONAL

    Inputs and outputs of the model

    • inputs Tensor, OPTIONAL

      Inputs of the model

    • outputs Tensor, OPTIONAL

      Outputs of the model

    • layers object, OPTIONAL

      Layers of the model

  • training object, OPTIONAL

    Training information of the model

    • git GitRepo, OPTIONAL

      Git repository of the training script

  • dataset object, OPTIONAL

    • git GitRepo, OPTIONAL

      Git repository of the dataset

Example

Here is an example model configuration JSON document:

{
   "created": "2015-10-31T22:22:56.015925234Z",
   "author": "Ce Gao <[email protected]>",
   "description": "CNN Model",
   "tags": [
       "cv"
   ],
   "labels": {
       "tensorflow.version": "2.0.0"
   },
   "framework": "TensorFlow",
   "format": "SavedModel",
   "size": 9223372036854775807,
   "metrics": { 
      "training": [
           {
               "name": "acc",
               "value": "0.928"
           }
      ],
   }
   "hyperParameters": [
       {
           "name": "batch_size",
           "value": "32"
       }
   ],
   "signature": {
       "inputs": [
           {
               "name": "input_1",
               "size": [
                   224,
                   224,
                   3
               ],
               "dType": "float64",
           }
       ],
       "outputs": [
           {
               "name": "output_1",
               "size": [
                   1,
                   1000
               ],
               "dType": "float64",
           }
       ],
       "layers": {
               "conv" : 1
       }
   },
   "training": {
       "git": {
           "repository": "[email protected]:kleveross/ormb.git",
           "revision": "22f1d8406d464b0c0874075539c1f2e96c253775"
       }
   },
   "dataset": {
       "git": {
           "repository": "[email protected]:kleveross/ormb.git",
           "revision": "22f1d8406d464b0c0874075539c1f2e96c253775"
       }
   }
}