Platform For AI:Model metrics

Limits

The serialized model metrics configuration cannot exceed 8,192 bytes in length.

Configure model metrics

When you register a new model, select Advanced Configuration to configure Model Metrics. For more information, see Register and manage models.

The following code shows a sample Model Metrics configuration.

{
  "Results": [{
    "Dataset": {
      "Uri": "oss://xxxx/"
    },
    "Metrics": {
      "lr": 0.000001,
      "train_loss": 2.6345
    }
  },{
    "Dataset": {
      "DatasetId": "d-alksdcjkasdfjhr"
    },
    "Metrics": {
      "lr": 0.000001,
      "train_loss": 2.6345
    }
  }]
}

The following table describes the key fields.

Field name		Required	Description
Results	Dataset	No	The dataset used to evaluate the model. The following reserved fields are supported: Uri: The path of the OSS bucket where the dataset is stored. DatasetId: The ID of the dataset. You can also customize `Dataset` parameters based on the sample configuration, such as by adding a dataset name.
	Metrics	Yes	The model performance evaluation metrics. You can also customize `Metrics` parameters based on the sample configuration: lr: The learning rate during model building. train_loss: The training loss.

View model metrics

After a model is registered, follow these steps to view the metrics for different model versions.

1. On the Model Management page, click the name of a model to go to its details page.

2. In the Model Version List section, click Details in the Operation column for the target model version.

3. In the Model Version panel, view the model metrics.

   Model metrics are displayed in the following two formats:

   • JSON

   • Table