ModelVersionProps

Properties for defining a ModelVersion.

See https://www.alibabacloud.com/help/ros/developer-reference/aliyun-pai-modelversion

Initializer

using AlibabaCloud.SDK.ROS.CDK.Pai;
new ModelVersionProps {
    object ModelId,
    object Uri,
    object ApprovalStatus = null,
    object ExtraInfo = null,
    object FormatType = null,
    object FrameworkType = null,
    object InferenceSpec = null,
    object Labels = null,
    object Metrics = null,
    object Options = null,
    object SourceId = null,
    object SourceType = null,
    object TrainingSpec = null,
    object VersionDescription = null,
    object VersionName = null
};

Properties

ModelId^Required

public object ModelId { get; set; }

• Type: object

Property modelId: Model ID.

For details about how to obtain the model ID.

Uri^Required

public object Uri { get; set; }

• Type: object

Property uri: Model version URI, that is, the location where the model is stored.

Possible values are:

• The HTTP(S) address of the model, for example: 'https://myweb.com/mymodel.tar.gz '.
• If the model is stored in OSS, the format is 'oss://./object '. For the endpoint configuration, see Access domain name and data center. For example, see OSS:// mybucket/'.

ApprovalStatus^Optional

public object ApprovalStatus { get; set; }

• Type: object

Property approvalStatus: Admission status, with values as follows: - Pending: to be determined.

• Approved: Allow to go online.
• Rejected: Online is not allowed.

ExtraInfo^Optional

public object ExtraInfo { get; set; }

• Type: object

Property extraInfo: The additional information.

FormatType^Optional

public object FormatType { get; set; }

• Type: object

Property formatType: Model format, possible values: - OfflineModel - SavedModel - Keras H5 - Frozen Pb - Caffe Prototxt - TorchScript - XGBoost - PMML - AlinkModel - ONNX.

FrameworkType^Optional

public object FrameworkType { get; set; }

• Type: object

Property frameworkType: Model framework, possible values: - Pytorch - XGBoost - Keras - Caffe - Alink - Xflow - TensorFlow.

InferenceSpec^Optional

public object InferenceSpec { get; set; }

• Type: object

Property inferenceSpec: Describes how to apply to downstream inference services.

For example, describe the processor and container of EAS. Example: { "processor": "tensorflow_gpu_1.12" }

Labels^Optional

public object Labels { get; set; }

• Type: object

Property labels: List of model version labels.

Metrics^Optional

public object Metrics { get; set; }

• Type: object

Property metrics: The metrics for the model.

The length after serialization is limited to 8,192. Example: { "Results": [{ "Dataset": { "DatasetId": "d-sdkjanksaklerhfd" }, "Metrics": { "cer": 0.175 } }, { "Dataset": { "Uri": "oss://xxxx/" }, "Metrics": { "cer": 0.172 } }] }

Options^Optional

public object Options { get; set; }

• Type: object

Property options: The extended field.

This is a JSON string.

SourceId^Optional

public object SourceId { get; set; }

• Type: object

Property sourceId: Source ID.

• When the source type is Custom, this field is not restricted.

• When the source is PAIFlow or TrainingService, the format is: region=,workspaceId=,kind=,id= Among them:

• region is the Alibaba Cloud region ID.

• workspaceId indicates the workspace ID.
• kind: is a type. Value: PipelineRun(PAIFlow workflow);ServiceJob (training service).
• id: is a unique identifier.

SourceType^Optional

public object SourceType { get; set; }

• Type: object

Property sourceType: Model source type, possible values: - Custom: Custom.

• PAIFlow:PAI workflow.
• TrainingService:PAI training service.

TrainingSpec^Optional

public object TrainingSpec { get; set; }

• Type: object

Property trainingSpec: Training configuration.

Configuration for fine-tuning, incremental training.

VersionDescription^Optional

public object VersionDescription { get; set; }

• Type: object

Property versionDescription: Model version description.

VersionName^Optional

public object VersionName { get; set; }

• Type: object

Property versionName: The model version, which is unique for each model.

If you leave this parameter empty, the first version is 0.1.0 by default. After that, the minor version number is increased by 1 in sequence. For example, the second version number is 0.2.0. A version number consists of a major version number, a minor version number, and a stage version number, separated by periods (.). The major version number and minor version number are numeric. The stage version number begins with a digit and can include numbers, underscores, and letters. For example, the version number is 1.1.0 or 2.3.4_beta.