---
title: "PATCH /projects/{+projectId}/datasets/{+datasetId}/models/{+modelId}"
method: PATCH
path: "/projects/{+projectId}/datasets/{+datasetId}/models/{+modelId}"
tags: ["models"]
---

# PATCH /projects/{+projectId}/datasets/{+datasetId}/models/{+modelId}

`PATCH /projects/{+projectId}/datasets/{+datasetId}/models/{+modelId}`

Patch specific fields in the specified model. # IAM Permissions Requires the `bigquery.models.updateMetadata` permission on the model.

## Path parameters

- `projectId` string, required
- `datasetId` string, required
- `modelId` string, required

## Request body

- Model
  - `defaultTrialId` string, int64 — Output only. The default trial_id to use in TVFs when the trial_id is not passed in. For single-objective [hyperparameter tuning](https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-hp-tuning-overview) models, this is the best trial ID. For multi-objective [hyperparameter tuning](https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-hp-tuning-overview) models, this is the smallest trial ID among all Pareto optimal trials.
  - `etag` string — Output only. A hash of this resource.
  - `description` string — Optional. A user-friendly description of this model.
  - `friendlyName` string — Optional. A descriptive name for this model.
  - `modelReference` ModelReference — Id path of a model.
    - `projectId` string — Required. The ID of the project containing this model.
    - `datasetId` string — Required. The ID of the dataset containing this model.
    - `modelId` string — Required. The ID of the model. The ID must contain only letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum length is 1,024 characters.
  - `hparamTrials` HparamTuningTrial[] — Output only. Trials of a [hyperparameter tuning](https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-hp-tuning-overview) model sorted by trial_id.
    - `trialId` string, int64 — 1-based index of the trial.
    - `evalLoss` number, double — Loss computed on the eval data at the end of trial.
    - `startTimeMs` string, int64 — Starting time of the trial.
    - `hparams` TrainingOptions — Options used in model training.
      - `autoClassWeights` boolean — Whether to calculate class weights automatically based on the popularity of each label.
      - `timeSeriesDataColumn` string — Column to be designated as time series data for ARIMA model.
      - `itemColumn` string — Item column specified for matrix factorization models.
      - `horizon` string, int64 — The number of periods ahead that need to be forecasted.
      - `subsample` number, double — Subsample fraction of the training data to grow tree to prevent overfitting for boosted tree models.
      - `dataSplitColumn` string — The column to split data with. This column won't be used as a feature. 1. When data_split_method is CUSTOM, the corresponding column should be boolean. The rows with true value tag are eval data, and the false are training data. 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION rows (from smallest to largest) in the corresponding column are used as training data, and the rest are eval data. It respects the order in Orderable data types: https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data_type_properties
      - `numTrials` string, int64 — Number of trials to run this hyperparameter tuning job.
      - `tfVersion` string — Based on the selected TF version, the corresponding docker image is used to train external models.
      - `modelUri` string — Google Cloud Storage URI from which the model was imported. Only applicable for imported models.
      - `holidayRegions` string[] — A list of geographical regions that are used for time series modeling.
      - `dropout` number, double — Dropout probability for dnn models.
      - `machineType` string — The type of the machine used to deploy and serve the model.
      - `minRelativeProgress` number, double — When early_stop is true, stops training when accuracy improvement is less than 'min_relative_progress'. Used only for iterative training algorithms.
      - `enableGlobalExplain` boolean — If true, enable global explanation during training.
      - `autoArimaMaxOrder` string, int64 — The max value of the sum of non-seasonal p and q.
      - `numParallelTree` string, int64 — Number of parallel trees constructed during each iteration for boosted tree models.
      - `forecastLimitLowerBound` number, double — The forecast limit lower bound that was used during ARIMA model training with limits. To see more details of the algorithm: https://otexts.com/fpp2/limits.html
      - `modelGardenModelName` string — The name of a Vertex model garden publisher model. Format is `publishers/{publisher}/models/{model}@{optional_version_id}`.
      - `boosterType` 'BOOSTER_TYPE_UNSPECIFIED' | 'GBTREE' | 'DART' — Booster type for boosted tree models.
      - `minReplicaCount` string, int64 — The minimum number of machine replicas that will be always deployed on an endpoint. This value must be greater than or equal to 1. The default value is 1.
      - `l2Regularization` number, double — L2 regularization coefficient.
      - `optimizationStrategy` 'OPTIMIZATION_STRATEGY_UNSPECIFIED' | 'BATCH_GRADIENT_DESCENT' | 'NORMAL_EQUATION' — Optimization strategy for training linear regression models.
      - `warmStart` boolean — Whether to train a model from the last checkpoint.
      - `numFactors` string, int64 — Num factors specified for matrix factorization models.
      - `colsampleBynode` number, double — Subsample ratio of columns for each node(split) for boosted tree models.
      - `instanceWeightColumn` string — Name of the instance weight column for training data. This column isn't be used as a feature.
      - `includeDrift` boolean — Include drift when fitting an ARIMA model.
      - `xgboostVersion` string — User-selected XGBoost versions for training of XGBoost models.
      - `l1Regularization` number, double — L1 regularization coefficient.
      - `feedbackType` 'FEEDBACK_TYPE_UNSPECIFIED' | 'IMPLICIT' | 'EXPLICIT' — Feedback type that specifies which algorithm to run for matrix factorization.
      - `huggingFaceModelId` string — The id of a Hugging Face model. For example, `google/gemma-2-2b-it`.
      - `minTreeChildWeight` string, int64 — Minimum sum of instance weight needed in a child for boosted tree models.
      - `colsampleBylevel` number, double — Subsample ratio of columns for each level for boosted tree models.
      - `maxParallelTrials` string, int64 — Maximum number of trials to run in parallel.
      - `dataSplitMethod` 'DATA_SPLIT_METHOD_UNSPECIFIED' | 'RANDOM' | 'CUSTOM' | 'SEQUENTIAL' | 'NO_SPLIT' | 'AUTO_SPLIT' — The data split type for training and evaluation, e.g. RANDOM.
      - `maxTimeSeriesLength` string, int64 — The maximum number of time points in a time series that can be used in modeling the trend component of the time series. Don't use this option with the `timeSeriesLengthFraction` or `minTimeSeriesLength` options.
      - `earlyStop` boolean — Whether to stop early when the loss doesn't improve significantly any more (compared to min_relative_progress). Used only for iterative training algorithms.
      - `hparamTuningObjectives` string[] — The target evaluation metrics to optimize the hyperparameters for.
      - `minTimeSeriesLength` string, int64 — The minimum number of time points in a time series that are used in modeling the trend component of the time series. If you use this option you must also set the `timeSeriesLengthFraction` option. This training option ensures that enough time points are available when you use `timeSeriesLengthFraction` in trend modeling. This is particularly important when forecasting multiple time series in a single query using `timeSeriesIdColumn`. If the total number of time points is less than the `minTimeSeriesLength` value, then the query uses all available time points.
      - `fitIntercept` boolean — Whether the model should include intercept during model training.
      - `calculatePValues` boolean — Whether or not p-value test should be computed for this model. Only available for linear and logistic regression models.
      - `pcaExplainedVarianceRatio` number, double — The minimum ratio of cumulative explained variance that needs to be given by the PCA model.
      - `reservationAffinityType` 'RESERVATION_AFFINITY_TYPE_UNSPECIFIED' | 'NO_RESERVATION' | 'ANY_RESERVATION' | 'SPECIFIC_RESERVATION' — Specifies the reservation affinity type used to configure a Vertex AI resource. The default value is `NO_RESERVATION`.
      - `categoryEncodingMethod` 'ENCODING_METHOD_UNSPECIFIED' | 'ONE_HOT_ENCODING' | 'LABEL_ENCODING' | 'DUMMY_ENCODING' — Categorical feature encoding method.
      - `holidayRegion` 'HOLIDAY_REGION_UNSPECIFIED' | 'GLOBAL' | 'NA' | 'JAPAC' | 'EMEA' | 'LAC' | 'AE' | 'AR' | 'AT' | 'AU' | 'BE' | 'BR' | 'CA' | 'CH' | 'CL' | 'CN' | 'CO' | 'CS' | 'CZ' | 'DE' | 'DK' | 'DZ' | 'EC' | 'EE' | 'EG' | 'ES' | 'FI' | 'FR' | 'GB' | 'GR' | 'HK' | 'HU' | 'ID' | 'IE' | 'IL' | 'IN' | 'IR' | 'IT' | 'JP' | 'KR' | 'LV' | 'MA' | 'MX' | 'MY' | 'NG' | 'NL' | 'NO' | 'NZ' | 'PE' | 'PH' | 'PK' | 'PL' | 'PT' | 'RO' | 'RS' | 'RU' | 'SA' | 'SE' | 'SG' | 'SI' | 'SK' | 'TH' | 'TR' | 'TW' | 'UA' | 'US' | 'VE' | 'VN' | 'ZA' — The geographical region based on which the holidays are considered in time series modeling. If a valid value is specified, then holiday effects modeling is enabled.
      - `standardizeFeatures` boolean — Whether to standardize numerical features. Default to true.
      - `maxReplicaCount` string, int64 — The maximum number of machine replicas that will be deployed on an endpoint. The default value is equal to min_replica_count.
      - `contributionMetric` string — The contribution metric. Applies to contribution analysis models. Allowed formats supported are for summable and summable ratio contribution metrics. These include expressions such as `SUM(x)` or `SUM(x)/SUM(y)`, where x and y are column names from the base table.
      - `integratedGradientsNumSteps` string, int64 — Number of integral steps for the integrated gradients explain method.
      - `isTestColumn` string — Name of the column used to determine the rows corresponding to control and test. Applies to contribution analysis models.
      - `vertexAiModelVersionAliases` string[] — The version aliases to apply in Vertex AI model registry. Always overwrite if the version aliases exists in a existing model.
      - `learnRateStrategy` 'LEARN_RATE_STRATEGY_UNSPECIFIED' | 'LINE_SEARCH' | 'CONSTANT' — The strategy to determine learn rate for the current iteration.
      - `minSplitLoss` number, double — Minimum split loss for boosted tree models.
      - `timeSeriesLengthFraction` number, double — The fraction of the interpolated length of the time series that's used to model the time series trend component. All of the time points of the time series are used to model the non-trend component. This training option accelerates modeling training without sacrificing much forecasting accuracy. You can use this option with `minTimeSeriesLength` but not with `maxTimeSeriesLength`.
      - `maxIterations` string, int64 — The maximum number of iterations in training. Used only for iterative training algorithms.
      - `decomposeTimeSeries` boolean — If true, perform decompose time series and save the results.
      - `dartNormalizeType` 'DART_NORMALIZE_TYPE_UNSPECIFIED' | 'TREE' | 'FOREST' — Type of normalization algorithm for boosted tree models using dart booster.
      - `numPrincipalComponents` string, int64 — Number of principal components to keep in the PCA model. Must be <= the number of features.
      - `activationFn` string — Activation function of the neural nets.
      - `approxGlobalFeatureContrib` boolean — Whether to use approximate feature contribution method in XGBoost model explanation for global explain.
      - `minAprioriSupport` number, double — The apriori support minimum. Applies to contribution analysis models.
      - `timeSeriesTimestampColumn` string — Column to be designated as time series timestamp for ARIMA model.
      - `l1RegActivation` number, double — L1 regularization coefficient to activations.
      - `dataSplitEvalFraction` number, double — The fraction of evaluation data over the whole input data. The rest of data will be used as training data. The format should be double. Accurate to two decimal places. Default value is 0.2.
      - `autoArimaMinOrder` string, int64 — The min value of the sum of non-seasonal p and q.
      - `scaleFeatures` boolean — If true, scale the feature values by dividing the feature standard deviation. Currently only apply to PCA.
      - `userColumn` string — User column specified for matrix factorization models.
      - `treeMethod` 'TREE_METHOD_UNSPECIFIED' | 'AUTO' | 'EXACT' | 'APPROX' | 'HIST' — Tree construction algorithm for boosted tree models.
      - `reservationAffinityValues` string[] — Corresponds to the label values of a reservation resource used by Vertex AI. This must be the full resource name of the reservation or reservation block.
      - `colsampleBytree` number, double — Subsample ratio of columns when constructing each tree for boosted tree models.
      - `kmeansInitializationMethod` 'KMEANS_INITIALIZATION_METHOD_UNSPECIFIED' | 'RANDOM' | 'CUSTOM' | 'KMEANS_PLUS_PLUS' — The method used to initialize the centroids for kmeans algorithm.
      - `optimizer` string — Optimizer used for training the neural nets.
      - `cleanSpikesAndDips` boolean — If true, clean spikes and dips in the input time series.
      - `forecastLimitUpperBound` number, double — The forecast limit upper bound that was used during ARIMA model training with limits.
      - `batchSize` string, int64 — Batch size for dnn models.
      - `inputLabelColumns` string[] — Name of input label columns in training data.
      - `endpointIdleTtl` string, google-duration — The idle TTL of the endpoint before the resources get destroyed. The default value is 6.5 hours.
      - `reservationAffinityKey` string — Corresponds to the label key of a reservation resource used by Vertex AI. To target a SPECIFIC_RESERVATION by name, use `compute.googleapis.com/reservation-name` as the key and specify the name of your reservation as its value.
      - `trendSmoothingWindowSize` string, int64 — Smoothing window size for the trend component. When a positive value is specified, a center moving average smoothing is applied on the history trend. When the smoothing window is out of the boundary at the beginning or the end of the trend, the first element or the last element is padded to fill the smoothing window before the average is applied.
      - `initialLearnRate` number, double — Specifies the initial learning rate for the line search learn rate strategy.
      - `numClusters` string, int64 — Number of clusters for clustering models.
      - `dataFrequency` 'DATA_FREQUENCY_UNSPECIFIED' | 'AUTO_FREQUENCY' | 'YEARLY' | 'QUARTERLY' | 'MONTHLY' | 'WEEKLY' | 'DAILY' | 'HOURLY' | 'PER_MINUTE' — The data frequency of a time series.
      - `pcaSolver` 'UNSPECIFIED' | 'FULL' | 'RANDOMIZED' | 'AUTO' — The solver for PCA.
      - `modelRegistry` 'MODEL_REGISTRY_UNSPECIFIED' | 'VERTEX_AI' — The model registry.
      - `walsAlpha` number, double — Hyperparameter for matrix factoration when implicit feedback type is specified.
      - `sampledShapleyNumPaths` string, int64 — Number of paths for the sampled Shapley explain method.
      - `colorSpace` 'COLOR_SPACE_UNSPECIFIED' | 'RGB' | 'HSV' | 'YIQ' | 'YUV' | 'GRAYSCALE' — Enums for color space, used for processing images in Object Table. See more details at https://www.tensorflow.org/io/tutorials/colorspace.
      - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
        - `p` string, int64 — Order of the autoregressive part.
        - `d` string, int64 — Order of the differencing part.
        - `q` string, int64 — Order of the moving-average part.
      - `lossType` 'LOSS_TYPE_UNSPECIFIED' | 'MEAN_SQUARED_LOSS' | 'MEAN_LOG_LOSS' — Type of loss function used during training run.
      - `timeSeriesIdColumns` string[] — The time series id columns that were used during ARIMA model training.
      - `distanceType` 'DISTANCE_TYPE_UNSPECIFIED' | 'EUCLIDEAN' | 'COSINE' — Distance type for clustering models.
      - `hiddenUnits` string[] — Hidden units for dnn models.
      - `maxTreeDepth` string, int64 — Maximum depth of a tree for boosted tree models.
      - `timeSeriesIdColumn` string — The time series id column that was used during ARIMA model training.
      - `kmeansInitializationColumn` string — The column used to provide the initial centroids for kmeans algorithm when kmeans_initialization_method is CUSTOM.
      - `labelClassWeights` object — Weights associated with each label class, for rebalancing the training data. Only applicable for classification models.
      - `dimensionIdColumns` string[] — Optional. Names of the columns to slice on. Applies to contribution analysis models.
      - `learnRate` number, double — Learning rate in training. Used only for iterative training algorithms.
      - `adjustStepChanges` boolean — If true, detect step changes and make data adjustment in the input time series.
      - `autoArima` boolean — Whether to enable auto ARIMA or not.
      - `budgetHours` number, double — Budget in hours for AutoML training.
    - `errorMessage` string — Error message for FAILED and INFEASIBLE trial.
    - `status` 'TRIAL_STATUS_UNSPECIFIED' | 'NOT_STARTED' | 'RUNNING' | 'SUCCEEDED' | 'FAILED' | 'INFEASIBLE' | 'STOPPED_EARLY' — The status of the trial.
    - `trainingLoss` number, double — Loss computed on the training data at the end of trial.
    - `evaluationMetrics` EvaluationMetrics — Evaluation metrics of a model. These are either computed on all training data or just the eval data based on whether eval data was used during training. These are not present for imported models.
      - `dimensionalityReductionMetrics` DimensionalityReductionMetrics — Model evaluation metrics for dimensionality reduction models.
        - `totalExplainedVarianceRatio` number, double — Total percentage of variance explained by the selected principal components.
      - `multiClassClassificationMetrics` MultiClassClassificationMetrics — Evaluation metrics for multi-class classification/classifier models.
        - `confusionMatrixList` ConfusionMatrix[] — Confusion matrix at different thresholds.
          - `confidenceThreshold` number, double — Confidence threshold used when computing the entries of the confusion matrix.
          - `rows` Row[] — One row per actual label.
            - `actualLabel` string — The original label of this row.
            - `entries` Entry[] — Info describing predicted label distribution.
              - …
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
      - `arimaForecastingMetrics` ArimaForecastingMetrics — Model evaluation metrics for ARIMA forecasting models.
        - `arimaFittingMetrics` ArimaFittingMetrics[] — Arima model fitting metrics.
          - `aic` number, double — AIC.
          - `logLikelihood` number, double — Log-likelihood.
          - `variance` number, double — Variance.
        - `hasDrift` boolean[] — Whether Arima model fitted with drift or not. It is always false when d is not 1.
        - `timeSeriesId` string[] — Id to differentiate different time series for the large-scale case.
        - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `arimaSingleModelForecastingMetrics` ArimaSingleModelForecastingMetrics[] — Repeated as there can be many metric sets (one for each model) in auto-arima and the large-scale case.
          - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
            - `p` string, int64 — Order of the autoregressive part.
            - `d` string, int64 — Order of the differencing part.
            - `q` string, int64 — Order of the moving-average part.
          - `hasDrift` boolean — Is arima model fitted with drift or not. It is always false when d is not 1.
          - `timeSeriesId` string — The time_series_id value for this time series. It will be one of the unique values from the time_series_id_column specified during ARIMA model training. Only present when time_series_id_column training option was used.
          - `timeSeriesIds` string[] — The tuple of time_series_ids identifying this time series. It will be one of the unique tuples of values present in the time_series_id_columns specified during ARIMA model training. Only present when time_series_id_columns training option was used and the order of values here are same as the order of time_series_id_columns.
          - `arimaFittingMetrics` ArimaFittingMetrics — ARIMA model fitting metrics.
            - `aic` number, double — AIC.
            - `logLikelihood` number, double — Log-likelihood.
            - `variance` number, double — Variance.
          - `hasStepChanges` boolean — If true, step_changes is a part of time series decomposition result.
          - `hasHolidayEffect` boolean — If true, holiday_effect is a part of time series decomposition result.
          - `hasSpikesAndDips` boolean — If true, spikes_and_dips is a part of time series decomposition result.
          - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `nonSeasonalOrder` ArimaOrder[] — Non-seasonal order.
          - `p` string, int64 — Order of the autoregressive part.
          - `d` string, int64 — Order of the differencing part.
          - `q` string, int64 — Order of the moving-average part.
      - `regressionMetrics` RegressionMetrics — Evaluation metrics for regression and explicit feedback type matrix factorization models.
        - `meanAbsoluteError` number, double — Mean absolute error.
        - `meanSquaredLogError` number, double — Mean squared log error.
        - `medianAbsoluteError` number, double — Median absolute error.
        - `meanSquaredError` number, double — Mean squared error.
        - `rSquared` number, double — R^2 score. This corresponds to r2_score in ML.EVALUATE.
      - `binaryClassificationMetrics` BinaryClassificationMetrics — Evaluation metrics for binary classification/classifier models.
        - `positiveLabel` string — Label representing the positive class.
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
        - `binaryConfusionMatrixList` BinaryConfusionMatrix[] — Binary confusion matrix at multiple thresholds.
          - `truePositives` string, int64 — Number of true samples predicted as true.
          - `falsePositives` string, int64 — Number of false samples predicted as true.
          - `precision` number, double — The fraction of actual positive predictions that had positive actual labels.
          - `recall` number, double — The fraction of actual positive labels that were given a positive prediction.
          - `accuracy` number, double — The fraction of predictions given the correct label.
          - `falseNegatives` string, int64 — Number of false samples predicted as false.
          - `positiveClassThreshold` number, double — Threshold value used when computing each of the following metric.
          - `f1Score` number, double — The equally weighted average of recall and precision.
          - `trueNegatives` string, int64 — Number of true samples predicted as false.
        - `negativeLabel` string — Label representing the negative class.
      - `clusteringMetrics` ClusteringMetrics — Evaluation metrics for clustering models.
        - `meanSquaredDistance` number, double — Mean of squared distances between each sample to its cluster centroid.
        - `daviesBouldinIndex` number, double — Davies-Bouldin index.
        - `clusters` Cluster[] — Information for all clusters.
          - `count` string, int64 — Count of training data rows that were assigned to this cluster.
          - `featureValues` FeatureValue[] — Values of highly variant features for this cluster.
            - `categoricalValue` CategoricalValue — Representative value of a categorical feature.
              - …
            - `featureColumn` string — The feature column name.
            - `numericalValue` number, double — The numerical feature value. This is the centroid value for this feature.
          - `centroidId` string, int64 — Centroid id.
      - `rankingMetrics` RankingMetrics — Evaluation metrics used by weighted-ALS models specified by feedback_type=implicit.
        - `averageRank` number, double — Determines the goodness of a ranking by computing the percentile rank from the predicted confidence and dividing it by the original rank.
        - `meanAveragePrecision` number, double — Calculates a precision per user for all the items by ranking them and then averages all the precisions across all the users.
        - `meanSquaredError` number, double — Similar to the mean squared error computed in regression and explicit recommendation models except instead of computing the rating directly, the output from evaluate is computed against a preference which is 1 or 0 depending on if the rating exists or not.
        - `normalizedDiscountedCumulativeGain` number, double — A metric to determine the goodness of a ranking calculated from the predicted confidence by comparing it to an ideal rank measured by the original ratings.
    - `endTimeMs` string, int64 — Ending time of the trial.
    - `hparamTuningEvaluationMetrics` EvaluationMetrics — Evaluation metrics of a model. These are either computed on all training data or just the eval data based on whether eval data was used during training. These are not present for imported models.
      - `dimensionalityReductionMetrics` DimensionalityReductionMetrics — Model evaluation metrics for dimensionality reduction models.
        - `totalExplainedVarianceRatio` number, double — Total percentage of variance explained by the selected principal components.
      - `multiClassClassificationMetrics` MultiClassClassificationMetrics — Evaluation metrics for multi-class classification/classifier models.
        - `confusionMatrixList` ConfusionMatrix[] — Confusion matrix at different thresholds.
          - `confidenceThreshold` number, double — Confidence threshold used when computing the entries of the confusion matrix.
          - `rows` Row[] — One row per actual label.
            - `actualLabel` string — The original label of this row.
            - `entries` Entry[] — Info describing predicted label distribution.
              - …
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
      - `arimaForecastingMetrics` ArimaForecastingMetrics — Model evaluation metrics for ARIMA forecasting models.
        - `arimaFittingMetrics` ArimaFittingMetrics[] — Arima model fitting metrics.
          - `aic` number, double — AIC.
          - `logLikelihood` number, double — Log-likelihood.
          - `variance` number, double — Variance.
        - `hasDrift` boolean[] — Whether Arima model fitted with drift or not. It is always false when d is not 1.
        - `timeSeriesId` string[] — Id to differentiate different time series for the large-scale case.
        - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `arimaSingleModelForecastingMetrics` ArimaSingleModelForecastingMetrics[] — Repeated as there can be many metric sets (one for each model) in auto-arima and the large-scale case.
          - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
            - `p` string, int64 — Order of the autoregressive part.
            - `d` string, int64 — Order of the differencing part.
            - `q` string, int64 — Order of the moving-average part.
          - `hasDrift` boolean — Is arima model fitted with drift or not. It is always false when d is not 1.
          - `timeSeriesId` string — The time_series_id value for this time series. It will be one of the unique values from the time_series_id_column specified during ARIMA model training. Only present when time_series_id_column training option was used.
          - `timeSeriesIds` string[] — The tuple of time_series_ids identifying this time series. It will be one of the unique tuples of values present in the time_series_id_columns specified during ARIMA model training. Only present when time_series_id_columns training option was used and the order of values here are same as the order of time_series_id_columns.
          - `arimaFittingMetrics` ArimaFittingMetrics — ARIMA model fitting metrics.
            - `aic` number, double — AIC.
            - `logLikelihood` number, double — Log-likelihood.
            - `variance` number, double — Variance.
          - `hasStepChanges` boolean — If true, step_changes is a part of time series decomposition result.
          - `hasHolidayEffect` boolean — If true, holiday_effect is a part of time series decomposition result.
          - `hasSpikesAndDips` boolean — If true, spikes_and_dips is a part of time series decomposition result.
          - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `nonSeasonalOrder` ArimaOrder[] — Non-seasonal order.
          - `p` string, int64 — Order of the autoregressive part.
          - `d` string, int64 — Order of the differencing part.
          - `q` string, int64 — Order of the moving-average part.
      - `regressionMetrics` RegressionMetrics — Evaluation metrics for regression and explicit feedback type matrix factorization models.
        - `meanAbsoluteError` number, double — Mean absolute error.
        - `meanSquaredLogError` number, double — Mean squared log error.
        - `medianAbsoluteError` number, double — Median absolute error.
        - `meanSquaredError` number, double — Mean squared error.
        - `rSquared` number, double — R^2 score. This corresponds to r2_score in ML.EVALUATE.
      - `binaryClassificationMetrics` BinaryClassificationMetrics — Evaluation metrics for binary classification/classifier models.
        - `positiveLabel` string — Label representing the positive class.
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
        - `binaryConfusionMatrixList` BinaryConfusionMatrix[] — Binary confusion matrix at multiple thresholds.
          - `truePositives` string, int64 — Number of true samples predicted as true.
          - `falsePositives` string, int64 — Number of false samples predicted as true.
          - `precision` number, double — The fraction of actual positive predictions that had positive actual labels.
          - `recall` number, double — The fraction of actual positive labels that were given a positive prediction.
          - `accuracy` number, double — The fraction of predictions given the correct label.
          - `falseNegatives` string, int64 — Number of false samples predicted as false.
          - `positiveClassThreshold` number, double — Threshold value used when computing each of the following metric.
          - `f1Score` number, double — The equally weighted average of recall and precision.
          - `trueNegatives` string, int64 — Number of true samples predicted as false.
        - `negativeLabel` string — Label representing the negative class.
      - `clusteringMetrics` ClusteringMetrics — Evaluation metrics for clustering models.
        - `meanSquaredDistance` number, double — Mean of squared distances between each sample to its cluster centroid.
        - `daviesBouldinIndex` number, double — Davies-Bouldin index.
        - `clusters` Cluster[] — Information for all clusters.
          - `count` string, int64 — Count of training data rows that were assigned to this cluster.
          - `featureValues` FeatureValue[] — Values of highly variant features for this cluster.
            - `categoricalValue` CategoricalValue — Representative value of a categorical feature.
              - …
            - `featureColumn` string — The feature column name.
            - `numericalValue` number, double — The numerical feature value. This is the centroid value for this feature.
          - `centroidId` string, int64 — Centroid id.
      - `rankingMetrics` RankingMetrics — Evaluation metrics used by weighted-ALS models specified by feedback_type=implicit.
        - `averageRank` number, double — Determines the goodness of a ranking by computing the percentile rank from the predicted confidence and dividing it by the original rank.
        - `meanAveragePrecision` number, double — Calculates a precision per user for all the items by ranking them and then averages all the precisions across all the users.
        - `meanSquaredError` number, double — Similar to the mean squared error computed in regression and explicit recommendation models except instead of computing the rating directly, the output from evaluate is computed against a preference which is 1 or 0 depending on if the rating exists or not.
        - `normalizedDiscountedCumulativeGain` number, double — A metric to determine the goodness of a ranking calculated from the predicted confidence by comparing it to an ideal rank measured by the original ratings.
  - `location` string — Output only. The geographic location where the model resides. This value is inherited from the dataset.
  - `labelColumns` StandardSqlField[] — Output only. Label columns that were used to train this model. The output of the model will have a "predicted_" prefix to these columns.
    - `name` string — Optional. The name of this field. Can be absent for struct fields.
    - `type` StandardSqlDataType — The data type of a variable such as a function argument. Examples include: * INT64: `{"typeKind": "INT64"}` * ARRAY: { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "STRING"} } * STRUCT>: { "typeKind": "STRUCT", "structType": { "fields": [ { "name": "x", "type": {"typeKind": "STRING"} }, { "name": "y", "type": { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "DATE"} } } ] } } * RANGE: { "typeKind": "RANGE", "rangeElementType": {"typeKind": "DATE"} }
      - `arrayElementType` StandardSqlDataType — recursive
      - `rangeElementType` StandardSqlDataType — recursive
      - `structType` StandardSqlStructType — The representation of a SQL STRUCT type.
        - `fields` StandardSqlField[] — Fields within the struct.
      - `typeKind` 'TYPE_KIND_UNSPECIFIED' | 'INT64' | 'BOOL' | 'FLOAT64' | 'STRING' | 'BYTES' | 'TIMESTAMP' | 'DATE' | 'TIME' | 'DATETIME' | 'INTERVAL' | 'GEOGRAPHY' | 'NUMERIC' | 'BIGNUMERIC' | 'JSON' | 'ARRAY' | 'STRUCT' | 'RANGE' — Required. The top level type of this field. Can be any GoogleSQL data type (e.g., "INT64", "DATE", "ARRAY").
  - `labels` object — The labels associated with this model. You can use these to organize and group your models. Label keys and values can be no longer than 63 characters, can only contain lowercase letters, numeric characters, underscores and dashes. International characters are allowed. Label values are optional. Label keys must start with a letter and each label in the list must have a different key.
  - `trainingRuns` TrainingRun[] — Information for all training runs in increasing order of start_time.
    - `modelLevelGlobalExplanation` GlobalExplanation — Global explanations containing the top most important features after training.
      - `classLabel` string — Class label for this set of global explanations. Will be empty/null for binary logistic and linear regression models. Sorted alphabetically in descending order.
      - `explanations` Explanation[] — A list of the top global explanations. Sorted by absolute value of attribution in descending order.
        - `attribution` number, double — Attribution of feature.
        - `featureName` string — The full feature name. For non-numerical features, will be formatted like `.`. Overall size of feature name will always be truncated to first 120 characters.
    - `vertexAiModelId` string — The model id in the [Vertex AI Model Registry](https://cloud.google.com/vertex-ai/docs/model-registry/introduction) for this training run.
    - `evaluationMetrics` EvaluationMetrics — Evaluation metrics of a model. These are either computed on all training data or just the eval data based on whether eval data was used during training. These are not present for imported models.
      - `dimensionalityReductionMetrics` DimensionalityReductionMetrics — Model evaluation metrics for dimensionality reduction models.
        - `totalExplainedVarianceRatio` number, double — Total percentage of variance explained by the selected principal components.
      - `multiClassClassificationMetrics` MultiClassClassificationMetrics — Evaluation metrics for multi-class classification/classifier models.
        - `confusionMatrixList` ConfusionMatrix[] — Confusion matrix at different thresholds.
          - `confidenceThreshold` number, double — Confidence threshold used when computing the entries of the confusion matrix.
          - `rows` Row[] — One row per actual label.
            - `actualLabel` string — The original label of this row.
            - `entries` Entry[] — Info describing predicted label distribution.
              - …
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
      - `arimaForecastingMetrics` ArimaForecastingMetrics — Model evaluation metrics for ARIMA forecasting models.
        - `arimaFittingMetrics` ArimaFittingMetrics[] — Arima model fitting metrics.
          - `aic` number, double — AIC.
          - `logLikelihood` number, double — Log-likelihood.
          - `variance` number, double — Variance.
        - `hasDrift` boolean[] — Whether Arima model fitted with drift or not. It is always false when d is not 1.
        - `timeSeriesId` string[] — Id to differentiate different time series for the large-scale case.
        - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `arimaSingleModelForecastingMetrics` ArimaSingleModelForecastingMetrics[] — Repeated as there can be many metric sets (one for each model) in auto-arima and the large-scale case.
          - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
            - `p` string, int64 — Order of the autoregressive part.
            - `d` string, int64 — Order of the differencing part.
            - `q` string, int64 — Order of the moving-average part.
          - `hasDrift` boolean — Is arima model fitted with drift or not. It is always false when d is not 1.
          - `timeSeriesId` string — The time_series_id value for this time series. It will be one of the unique values from the time_series_id_column specified during ARIMA model training. Only present when time_series_id_column training option was used.
          - `timeSeriesIds` string[] — The tuple of time_series_ids identifying this time series. It will be one of the unique tuples of values present in the time_series_id_columns specified during ARIMA model training. Only present when time_series_id_columns training option was used and the order of values here are same as the order of time_series_id_columns.
          - `arimaFittingMetrics` ArimaFittingMetrics — ARIMA model fitting metrics.
            - `aic` number, double — AIC.
            - `logLikelihood` number, double — Log-likelihood.
            - `variance` number, double — Variance.
          - `hasStepChanges` boolean — If true, step_changes is a part of time series decomposition result.
          - `hasHolidayEffect` boolean — If true, holiday_effect is a part of time series decomposition result.
          - `hasSpikesAndDips` boolean — If true, spikes_and_dips is a part of time series decomposition result.
          - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `nonSeasonalOrder` ArimaOrder[] — Non-seasonal order.
          - `p` string, int64 — Order of the autoregressive part.
          - `d` string, int64 — Order of the differencing part.
          - `q` string, int64 — Order of the moving-average part.
      - `regressionMetrics` RegressionMetrics — Evaluation metrics for regression and explicit feedback type matrix factorization models.
        - `meanAbsoluteError` number, double — Mean absolute error.
        - `meanSquaredLogError` number, double — Mean squared log error.
        - `medianAbsoluteError` number, double — Median absolute error.
        - `meanSquaredError` number, double — Mean squared error.
        - `rSquared` number, double — R^2 score. This corresponds to r2_score in ML.EVALUATE.
      - `binaryClassificationMetrics` BinaryClassificationMetrics — Evaluation metrics for binary classification/classifier models.
        - `positiveLabel` string — Label representing the positive class.
        - `aggregateClassificationMetrics` AggregateClassificationMetrics — Aggregate metrics for classification/classifier models. For multi-class models, the metrics are either macro-averaged or micro-averaged. When macro-averaged, the metrics are calculated for each label and then an unweighted average is taken of those values. When micro-averaged, the metric is calculated globally by counting the total number of correctly predicted rows.
          - `accuracy` number, double — Accuracy is the fraction of predictions given the correct label. For multiclass this is a micro-averaged metric.
          - `rocAuc` number, double — Area Under a ROC Curve. For multiclass this is a macro-averaged metric.
          - `threshold` number, double — Threshold at which the metrics are computed. For binary classification models this is the positive class threshold. For multi-class classification models this is the confidence threshold.
          - `logLoss` number, double — Logarithmic Loss. For multiclass this is a macro-averaged metric.
          - `precision` number, double — Precision is the fraction of actual positive predictions that had positive actual labels. For multiclass this is a macro-averaged metric treating each class as a binary classifier.
          - `recall` number, double — Recall is the fraction of actual positive labels that were given a positive prediction. For multiclass this is a macro-averaged metric.
          - `f1Score` number, double — The F1 score is an average of recall and precision. For multiclass this is a macro-averaged metric.
        - `binaryConfusionMatrixList` BinaryConfusionMatrix[] — Binary confusion matrix at multiple thresholds.
          - `truePositives` string, int64 — Number of true samples predicted as true.
          - `falsePositives` string, int64 — Number of false samples predicted as true.
          - `precision` number, double — The fraction of actual positive predictions that had positive actual labels.
          - `recall` number, double — The fraction of actual positive labels that were given a positive prediction.
          - `accuracy` number, double — The fraction of predictions given the correct label.
          - `falseNegatives` string, int64 — Number of false samples predicted as false.
          - `positiveClassThreshold` number, double — Threshold value used when computing each of the following metric.
          - `f1Score` number, double — The equally weighted average of recall and precision.
          - `trueNegatives` string, int64 — Number of true samples predicted as false.
        - `negativeLabel` string — Label representing the negative class.
      - `clusteringMetrics` ClusteringMetrics — Evaluation metrics for clustering models.
        - `meanSquaredDistance` number, double — Mean of squared distances between each sample to its cluster centroid.
        - `daviesBouldinIndex` number, double — Davies-Bouldin index.
        - `clusters` Cluster[] — Information for all clusters.
          - `count` string, int64 — Count of training data rows that were assigned to this cluster.
          - `featureValues` FeatureValue[] — Values of highly variant features for this cluster.
            - `categoricalValue` CategoricalValue — Representative value of a categorical feature.
              - …
            - `featureColumn` string — The feature column name.
            - `numericalValue` number, double — The numerical feature value. This is the centroid value for this feature.
          - `centroidId` string, int64 — Centroid id.
      - `rankingMetrics` RankingMetrics — Evaluation metrics used by weighted-ALS models specified by feedback_type=implicit.
        - `averageRank` number, double — Determines the goodness of a ranking by computing the percentile rank from the predicted confidence and dividing it by the original rank.
        - `meanAveragePrecision` number, double — Calculates a precision per user for all the items by ranking them and then averages all the precisions across all the users.
        - `meanSquaredError` number, double — Similar to the mean squared error computed in regression and explicit recommendation models except instead of computing the rating directly, the output from evaluate is computed against a preference which is 1 or 0 depending on if the rating exists or not.
        - `normalizedDiscountedCumulativeGain` number, double — A metric to determine the goodness of a ranking calculated from the predicted confidence by comparing it to an ideal rank measured by the original ratings.
    - `dataSplitResult` DataSplitResult — Data split result. This contains references to the training and evaluation data tables that were used to train the model.
      - `testTable` TableReference
        - `datasetId` string — Required. The ID of the dataset containing this table.
        - `projectId` string — Required. The ID of the project containing this table.
        - `tableId` string — Required. The ID of the table. The ID can contain Unicode characters in category L (letter), M (mark), N (number), Pc (connector, including underscore), Pd (dash), and Zs (space). For more information, see [General Category](https://wikipedia.org/wiki/Unicode_character_property#General_Category). The maximum length is 1,024 characters. Certain operations allow suffixing of the table ID with a partition decorator, such as `sample_table$20190123`.
      - `trainingTable` TableReference
        - `datasetId` string — Required. The ID of the dataset containing this table.
        - `projectId` string — Required. The ID of the project containing this table.
        - `tableId` string — Required. The ID of the table. The ID can contain Unicode characters in category L (letter), M (mark), N (number), Pc (connector, including underscore), Pd (dash), and Zs (space). For more information, see [General Category](https://wikipedia.org/wiki/Unicode_character_property#General_Category). The maximum length is 1,024 characters. Certain operations allow suffixing of the table ID with a partition decorator, such as `sample_table$20190123`.
      - `evaluationTable` TableReference
        - `datasetId` string — Required. The ID of the dataset containing this table.
        - `projectId` string — Required. The ID of the project containing this table.
        - `tableId` string — Required. The ID of the table. The ID can contain Unicode characters in category L (letter), M (mark), N (number), Pc (connector, including underscore), Pd (dash), and Zs (space). For more information, see [General Category](https://wikipedia.org/wiki/Unicode_character_property#General_Category). The maximum length is 1,024 characters. Certain operations allow suffixing of the table ID with a partition decorator, such as `sample_table$20190123`.
    - `classLevelGlobalExplanations` GlobalExplanation[] — Output only. Global explanation contains the explanation of top features on the class level. Applies to classification models only.
      - `classLabel` string — Class label for this set of global explanations. Will be empty/null for binary logistic and linear regression models. Sorted alphabetically in descending order.
      - `explanations` Explanation[] — A list of the top global explanations. Sorted by absolute value of attribution in descending order.
        - `attribution` number, double — Attribution of feature.
        - `featureName` string — The full feature name. For non-numerical features, will be formatted like `.`. Overall size of feature name will always be truncated to first 120 characters.
    - `trainingStartTime` string, int64 — Output only. The start time of this training run, in milliseconds since epoch.
    - `vertexAiModelVersion` string — Output only. The model version in the [Vertex AI Model Registry](https://cloud.google.com/vertex-ai/docs/model-registry/introduction) for this training run.
    - `trainingOptions` TrainingOptions — Options used in model training.
      - `autoClassWeights` boolean — Whether to calculate class weights automatically based on the popularity of each label.
      - `timeSeriesDataColumn` string — Column to be designated as time series data for ARIMA model.
      - `itemColumn` string — Item column specified for matrix factorization models.
      - `horizon` string, int64 — The number of periods ahead that need to be forecasted.
      - `subsample` number, double — Subsample fraction of the training data to grow tree to prevent overfitting for boosted tree models.
      - `dataSplitColumn` string — The column to split data with. This column won't be used as a feature. 1. When data_split_method is CUSTOM, the corresponding column should be boolean. The rows with true value tag are eval data, and the false are training data. 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION rows (from smallest to largest) in the corresponding column are used as training data, and the rest are eval data. It respects the order in Orderable data types: https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data_type_properties
      - `numTrials` string, int64 — Number of trials to run this hyperparameter tuning job.
      - `tfVersion` string — Based on the selected TF version, the corresponding docker image is used to train external models.
      - `modelUri` string — Google Cloud Storage URI from which the model was imported. Only applicable for imported models.
      - `holidayRegions` string[] — A list of geographical regions that are used for time series modeling.
      - `dropout` number, double — Dropout probability for dnn models.
      - `machineType` string — The type of the machine used to deploy and serve the model.
      - `minRelativeProgress` number, double — When early_stop is true, stops training when accuracy improvement is less than 'min_relative_progress'. Used only for iterative training algorithms.
      - `enableGlobalExplain` boolean — If true, enable global explanation during training.
      - `autoArimaMaxOrder` string, int64 — The max value of the sum of non-seasonal p and q.
      - `numParallelTree` string, int64 — Number of parallel trees constructed during each iteration for boosted tree models.
      - `forecastLimitLowerBound` number, double — The forecast limit lower bound that was used during ARIMA model training with limits. To see more details of the algorithm: https://otexts.com/fpp2/limits.html
      - `modelGardenModelName` string — The name of a Vertex model garden publisher model. Format is `publishers/{publisher}/models/{model}@{optional_version_id}`.
      - `boosterType` 'BOOSTER_TYPE_UNSPECIFIED' | 'GBTREE' | 'DART' — Booster type for boosted tree models.
      - `minReplicaCount` string, int64 — The minimum number of machine replicas that will be always deployed on an endpoint. This value must be greater than or equal to 1. The default value is 1.
      - `l2Regularization` number, double — L2 regularization coefficient.
      - `optimizationStrategy` 'OPTIMIZATION_STRATEGY_UNSPECIFIED' | 'BATCH_GRADIENT_DESCENT' | 'NORMAL_EQUATION' — Optimization strategy for training linear regression models.
      - `warmStart` boolean — Whether to train a model from the last checkpoint.
      - `numFactors` string, int64 — Num factors specified for matrix factorization models.
      - `colsampleBynode` number, double — Subsample ratio of columns for each node(split) for boosted tree models.
      - `instanceWeightColumn` string — Name of the instance weight column for training data. This column isn't be used as a feature.
      - `includeDrift` boolean — Include drift when fitting an ARIMA model.
      - `xgboostVersion` string — User-selected XGBoost versions for training of XGBoost models.
      - `l1Regularization` number, double — L1 regularization coefficient.
      - `feedbackType` 'FEEDBACK_TYPE_UNSPECIFIED' | 'IMPLICIT' | 'EXPLICIT' — Feedback type that specifies which algorithm to run for matrix factorization.
      - `huggingFaceModelId` string — The id of a Hugging Face model. For example, `google/gemma-2-2b-it`.
      - `minTreeChildWeight` string, int64 — Minimum sum of instance weight needed in a child for boosted tree models.
      - `colsampleBylevel` number, double — Subsample ratio of columns for each level for boosted tree models.
      - `maxParallelTrials` string, int64 — Maximum number of trials to run in parallel.
      - `dataSplitMethod` 'DATA_SPLIT_METHOD_UNSPECIFIED' | 'RANDOM' | 'CUSTOM' | 'SEQUENTIAL' | 'NO_SPLIT' | 'AUTO_SPLIT' — The data split type for training and evaluation, e.g. RANDOM.
      - `maxTimeSeriesLength` string, int64 — The maximum number of time points in a time series that can be used in modeling the trend component of the time series. Don't use this option with the `timeSeriesLengthFraction` or `minTimeSeriesLength` options.
      - `earlyStop` boolean — Whether to stop early when the loss doesn't improve significantly any more (compared to min_relative_progress). Used only for iterative training algorithms.
      - `hparamTuningObjectives` string[] — The target evaluation metrics to optimize the hyperparameters for.
      - `minTimeSeriesLength` string, int64 — The minimum number of time points in a time series that are used in modeling the trend component of the time series. If you use this option you must also set the `timeSeriesLengthFraction` option. This training option ensures that enough time points are available when you use `timeSeriesLengthFraction` in trend modeling. This is particularly important when forecasting multiple time series in a single query using `timeSeriesIdColumn`. If the total number of time points is less than the `minTimeSeriesLength` value, then the query uses all available time points.
      - `fitIntercept` boolean — Whether the model should include intercept during model training.
      - `calculatePValues` boolean — Whether or not p-value test should be computed for this model. Only available for linear and logistic regression models.
      - `pcaExplainedVarianceRatio` number, double — The minimum ratio of cumulative explained variance that needs to be given by the PCA model.
      - `reservationAffinityType` 'RESERVATION_AFFINITY_TYPE_UNSPECIFIED' | 'NO_RESERVATION' | 'ANY_RESERVATION' | 'SPECIFIC_RESERVATION' — Specifies the reservation affinity type used to configure a Vertex AI resource. The default value is `NO_RESERVATION`.
      - `categoryEncodingMethod` 'ENCODING_METHOD_UNSPECIFIED' | 'ONE_HOT_ENCODING' | 'LABEL_ENCODING' | 'DUMMY_ENCODING' — Categorical feature encoding method.
      - `holidayRegion` 'HOLIDAY_REGION_UNSPECIFIED' | 'GLOBAL' | 'NA' | 'JAPAC' | 'EMEA' | 'LAC' | 'AE' | 'AR' | 'AT' | 'AU' | 'BE' | 'BR' | 'CA' | 'CH' | 'CL' | 'CN' | 'CO' | 'CS' | 'CZ' | 'DE' | 'DK' | 'DZ' | 'EC' | 'EE' | 'EG' | 'ES' | 'FI' | 'FR' | 'GB' | 'GR' | 'HK' | 'HU' | 'ID' | 'IE' | 'IL' | 'IN' | 'IR' | 'IT' | 'JP' | 'KR' | 'LV' | 'MA' | 'MX' | 'MY' | 'NG' | 'NL' | 'NO' | 'NZ' | 'PE' | 'PH' | 'PK' | 'PL' | 'PT' | 'RO' | 'RS' | 'RU' | 'SA' | 'SE' | 'SG' | 'SI' | 'SK' | 'TH' | 'TR' | 'TW' | 'UA' | 'US' | 'VE' | 'VN' | 'ZA' — The geographical region based on which the holidays are considered in time series modeling. If a valid value is specified, then holiday effects modeling is enabled.
      - `standardizeFeatures` boolean — Whether to standardize numerical features. Default to true.
      - `maxReplicaCount` string, int64 — The maximum number of machine replicas that will be deployed on an endpoint. The default value is equal to min_replica_count.
      - `contributionMetric` string — The contribution metric. Applies to contribution analysis models. Allowed formats supported are for summable and summable ratio contribution metrics. These include expressions such as `SUM(x)` or `SUM(x)/SUM(y)`, where x and y are column names from the base table.
      - `integratedGradientsNumSteps` string, int64 — Number of integral steps for the integrated gradients explain method.
      - `isTestColumn` string — Name of the column used to determine the rows corresponding to control and test. Applies to contribution analysis models.
      - `vertexAiModelVersionAliases` string[] — The version aliases to apply in Vertex AI model registry. Always overwrite if the version aliases exists in a existing model.
      - `learnRateStrategy` 'LEARN_RATE_STRATEGY_UNSPECIFIED' | 'LINE_SEARCH' | 'CONSTANT' — The strategy to determine learn rate for the current iteration.
      - `minSplitLoss` number, double — Minimum split loss for boosted tree models.
      - `timeSeriesLengthFraction` number, double — The fraction of the interpolated length of the time series that's used to model the time series trend component. All of the time points of the time series are used to model the non-trend component. This training option accelerates modeling training without sacrificing much forecasting accuracy. You can use this option with `minTimeSeriesLength` but not with `maxTimeSeriesLength`.
      - `maxIterations` string, int64 — The maximum number of iterations in training. Used only for iterative training algorithms.
      - `decomposeTimeSeries` boolean — If true, perform decompose time series and save the results.
      - `dartNormalizeType` 'DART_NORMALIZE_TYPE_UNSPECIFIED' | 'TREE' | 'FOREST' — Type of normalization algorithm for boosted tree models using dart booster.
      - `numPrincipalComponents` string, int64 — Number of principal components to keep in the PCA model. Must be <= the number of features.
      - `activationFn` string — Activation function of the neural nets.
      - `approxGlobalFeatureContrib` boolean — Whether to use approximate feature contribution method in XGBoost model explanation for global explain.
      - `minAprioriSupport` number, double — The apriori support minimum. Applies to contribution analysis models.
      - `timeSeriesTimestampColumn` string — Column to be designated as time series timestamp for ARIMA model.
      - `l1RegActivation` number, double — L1 regularization coefficient to activations.
      - `dataSplitEvalFraction` number, double — The fraction of evaluation data over the whole input data. The rest of data will be used as training data. The format should be double. Accurate to two decimal places. Default value is 0.2.
      - `autoArimaMinOrder` string, int64 — The min value of the sum of non-seasonal p and q.
      - `scaleFeatures` boolean — If true, scale the feature values by dividing the feature standard deviation. Currently only apply to PCA.
      - `userColumn` string — User column specified for matrix factorization models.
      - `treeMethod` 'TREE_METHOD_UNSPECIFIED' | 'AUTO' | 'EXACT' | 'APPROX' | 'HIST' — Tree construction algorithm for boosted tree models.
      - `reservationAffinityValues` string[] — Corresponds to the label values of a reservation resource used by Vertex AI. This must be the full resource name of the reservation or reservation block.
      - `colsampleBytree` number, double — Subsample ratio of columns when constructing each tree for boosted tree models.
      - `kmeansInitializationMethod` 'KMEANS_INITIALIZATION_METHOD_UNSPECIFIED' | 'RANDOM' | 'CUSTOM' | 'KMEANS_PLUS_PLUS' — The method used to initialize the centroids for kmeans algorithm.
      - `optimizer` string — Optimizer used for training the neural nets.
      - `cleanSpikesAndDips` boolean — If true, clean spikes and dips in the input time series.
      - `forecastLimitUpperBound` number, double — The forecast limit upper bound that was used during ARIMA model training with limits.
      - `batchSize` string, int64 — Batch size for dnn models.
      - `inputLabelColumns` string[] — Name of input label columns in training data.
      - `endpointIdleTtl` string, google-duration — The idle TTL of the endpoint before the resources get destroyed. The default value is 6.5 hours.
      - `reservationAffinityKey` string — Corresponds to the label key of a reservation resource used by Vertex AI. To target a SPECIFIC_RESERVATION by name, use `compute.googleapis.com/reservation-name` as the key and specify the name of your reservation as its value.
      - `trendSmoothingWindowSize` string, int64 — Smoothing window size for the trend component. When a positive value is specified, a center moving average smoothing is applied on the history trend. When the smoothing window is out of the boundary at the beginning or the end of the trend, the first element or the last element is padded to fill the smoothing window before the average is applied.
      - `initialLearnRate` number, double — Specifies the initial learning rate for the line search learn rate strategy.
      - `numClusters` string, int64 — Number of clusters for clustering models.
      - `dataFrequency` 'DATA_FREQUENCY_UNSPECIFIED' | 'AUTO_FREQUENCY' | 'YEARLY' | 'QUARTERLY' | 'MONTHLY' | 'WEEKLY' | 'DAILY' | 'HOURLY' | 'PER_MINUTE' — The data frequency of a time series.
      - `pcaSolver` 'UNSPECIFIED' | 'FULL' | 'RANDOMIZED' | 'AUTO' — The solver for PCA.
      - `modelRegistry` 'MODEL_REGISTRY_UNSPECIFIED' | 'VERTEX_AI' — The model registry.
      - `walsAlpha` number, double — Hyperparameter for matrix factoration when implicit feedback type is specified.
      - `sampledShapleyNumPaths` string, int64 — Number of paths for the sampled Shapley explain method.
      - `colorSpace` 'COLOR_SPACE_UNSPECIFIED' | 'RGB' | 'HSV' | 'YIQ' | 'YUV' | 'GRAYSCALE' — Enums for color space, used for processing images in Object Table. See more details at https://www.tensorflow.org/io/tutorials/colorspace.
      - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
        - `p` string, int64 — Order of the autoregressive part.
        - `d` string, int64 — Order of the differencing part.
        - `q` string, int64 — Order of the moving-average part.
      - `lossType` 'LOSS_TYPE_UNSPECIFIED' | 'MEAN_SQUARED_LOSS' | 'MEAN_LOG_LOSS' — Type of loss function used during training run.
      - `timeSeriesIdColumns` string[] — The time series id columns that were used during ARIMA model training.
      - `distanceType` 'DISTANCE_TYPE_UNSPECIFIED' | 'EUCLIDEAN' | 'COSINE' — Distance type for clustering models.
      - `hiddenUnits` string[] — Hidden units for dnn models.
      - `maxTreeDepth` string, int64 — Maximum depth of a tree for boosted tree models.
      - `timeSeriesIdColumn` string — The time series id column that was used during ARIMA model training.
      - `kmeansInitializationColumn` string — The column used to provide the initial centroids for kmeans algorithm when kmeans_initialization_method is CUSTOM.
      - `labelClassWeights` object — Weights associated with each label class, for rebalancing the training data. Only applicable for classification models.
      - `dimensionIdColumns` string[] — Optional. Names of the columns to slice on. Applies to contribution analysis models.
      - `learnRate` number, double — Learning rate in training. Used only for iterative training algorithms.
      - `adjustStepChanges` boolean — If true, detect step changes and make data adjustment in the input time series.
      - `autoArima` boolean — Whether to enable auto ARIMA or not.
      - `budgetHours` number, double — Budget in hours for AutoML training.
    - `results` IterationResult[] — Output only. Output of each iteration run, results.size() <= max_iterations.
      - `evalLoss` number, double — Loss computed on the eval data at the end of iteration.
      - `index` integer — Index of the iteration, 0 based.
      - `arimaResult` ArimaResult — (Auto-)arima fitting result. Wrap everything in ArimaResult for easier refactoring if we want to use model-specific iteration results.
        - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
        - `arimaModelInfo` ArimaModelInfo[] — This message is repeated because there are multiple arima models fitted in auto-arima. For non-auto-arima model, its size is one.
          - `hasDrift` boolean — Whether Arima model fitted with drift or not. It is always false when d is not 1.
          - `timeSeriesId` string — The time_series_id value for this time series. It will be one of the unique values from the time_series_id_column specified during ARIMA model training. Only present when time_series_id_column training option was used.
          - `arimaCoefficients` ArimaCoefficients — Arima coefficients.
            - `autoRegressiveCoefficients` number[] — Auto-regressive coefficients, an array of double.
            - `interceptCoefficient` number, double — Intercept coefficient, just a double not an array.
            - `movingAverageCoefficients` number[] — Moving-average coefficients, an array of double.
          - `nonSeasonalOrder` ArimaOrder — Arima order, can be used for both non-seasonal and seasonal parts.
            - `p` string, int64 — Order of the autoregressive part.
            - `d` string, int64 — Order of the differencing part.
            - `q` string, int64 — Order of the moving-average part.
          - `arimaFittingMetrics` ArimaFittingMetrics — ARIMA model fitting metrics.
            - `aic` number, double — AIC.
            - `logLikelihood` number, double — Log-likelihood.
            - `variance` number, double — Variance.
          - `hasStepChanges` boolean — If true, step_changes is a part of time series decomposition result.
          - `hasHolidayEffect` boolean — If true, holiday_effect is a part of time series decomposition result.
          - `hasSpikesAndDips` boolean — If true, spikes_and_dips is a part of time series decomposition result.
          - `seasonalPeriods` string[] — Seasonal periods. Repeated because multiple periods are supported for one time series.
          - `timeSeriesIds` string[] — The tuple of time_series_ids identifying this time series. It will be one of the unique tuples of values present in the time_series_id_columns specified during ARIMA model training. Only present when time_series_id_columns training option was used and the order of values here are same as the order of time_series_id_columns.
      - `learnRate` number, double — Learn rate used for this iteration.
      - `durationMs` string, int64 — Time taken to run the iteration in milliseconds.
      - `principalComponentInfos` PrincipalComponentInfo[] — The information of the principal components.
        - `cumulativeExplainedVarianceRatio` number, double — The explained_variance is pre-ordered in the descending order to compute the cumulative explained variance ratio.
        - `principalComponentId` string, int64 — Id of the principal component.
        - `explainedVariance` number, double — Explained variance by this principal component, which is simply the eigenvalue.
        - `explainedVarianceRatio` number, double — Explained_variance over the total explained variance.
      - `trainingLoss` number, double — Loss computed on the training data at the end of iteration.
      - `clusterInfos` ClusterInfo[] — Information about top clusters for clustering models.
        - `centroidId` string, int64 — Centroid id.
        - `clusterRadius` number, double — Cluster radius, the average distance from centroid to each point assigned to the cluster.
        - `clusterSize` string, int64 — Cluster size, the total number of points assigned to the cluster.
    - `startTime` string, google-datetime — Output only. The start time of this training run.
  - `creationTime` string, int64 — Output only. The time when this model was created, in millisecs since the epoch.
  - `expirationTime` string, int64 — Optional. The time when this model expires, in milliseconds since the epoch. If not present, the model will persist indefinitely. Expired models will be deleted and their storage reclaimed. The defaultTableExpirationMs property of the encapsulating dataset can be used to set a default expirationTime on newly created models.
  - `hparamSearchSpaces` HparamSearchSpaces — Hyperparameter search spaces. These should be a subset of training_options.
    - `dartNormalizeType` StringHparamSearchSpace — Search space for string and enum.
      - `candidates` string[] — Canididates for the string or enum parameter in lower case.
    - `numClusters` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
    - `optimizer` StringHparamSearchSpace — Search space for string and enum.
      - `candidates` string[] — Canididates for the string or enum parameter in lower case.
    - `hiddenUnits` IntArrayHparamSearchSpace — Search space for int array.
      - `candidates` IntArray[] — Candidates for the int array parameter.
        - `elements` string[] — Elements in the int array.
    - `maxTreeDepth` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
    - `colsampleBytree` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `subsample` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `l1Reg` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `colsampleBynode` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `minSplitLoss` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `numFactors` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
    - `treeMethod` StringHparamSearchSpace — Search space for string and enum.
      - `candidates` string[] — Canididates for the string or enum parameter in lower case.
    - `l2Reg` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `colsampleBylevel` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `dropout` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `minTreeChildWeight` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
    - `boosterType` StringHparamSearchSpace — Search space for string and enum.
      - `candidates` string[] — Canididates for the string or enum parameter in lower case.
    - `batchSize` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
    - `activationFn` StringHparamSearchSpace — Search space for string and enum.
      - `candidates` string[] — Canididates for the string or enum parameter in lower case.
    - `learnRate` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `walsAlpha` DoubleHparamSearchSpace — Search space for a double hyperparameter.
      - `candidates` DoubleCandidates — Discrete candidates of a double hyperparameter.
        - `candidates` number[] — Candidates for the double parameter in increasing order.
      - `range` DoubleRange — Range of a double hyperparameter.
        - `min` number, double — Min value of the double parameter.
        - `max` number, double — Max value of the double parameter.
    - `numParallelTree` IntHparamSearchSpace — Search space for an int hyperparameter.
      - `range` IntRange — Range of an int hyperparameter.
        - `min` string, int64 — Min value of the int parameter.
        - `max` string, int64 — Max value of the int parameter.
      - `candidates` IntCandidates — Discrete candidates of an int hyperparameter.
        - `candidates` string[] — Candidates for the int parameter in increasing order.
  - `encryptionConfiguration` EncryptionConfiguration — Configuration for Cloud KMS encryption settings.
    - `kmsKeyName` string — Optional. Describes the Cloud KMS encryption key that will be used to protect destination BigQuery table. The BigQuery Service Account associated with your project requires access to this encryption key.
  - `featureColumns` StandardSqlField[] — Output only. Input feature columns for the model inference. If the model is trained with TRANSFORM clause, these are the input of the TRANSFORM clause.
    - `name` string — Optional. The name of this field. Can be absent for struct fields.
    - `type` StandardSqlDataType — The data type of a variable such as a function argument. Examples include: * INT64: `{"typeKind": "INT64"}` * ARRAY: { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "STRING"} } * STRUCT>: { "typeKind": "STRUCT", "structType": { "fields": [ { "name": "x", "type": {"typeKind": "STRING"} }, { "name": "y", "type": { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "DATE"} } } ] } } * RANGE: { "typeKind": "RANGE", "rangeElementType": {"typeKind": "DATE"} }
      - `arrayElementType` StandardSqlDataType — recursive
      - `rangeElementType` StandardSqlDataType — recursive
      - `structType` StandardSqlStructType — The representation of a SQL STRUCT type.
        - `fields` StandardSqlField[] — Fields within the struct.
      - `typeKind` 'TYPE_KIND_UNSPECIFIED' | 'INT64' | 'BOOL' | 'FLOAT64' | 'STRING' | 'BYTES' | 'TIMESTAMP' | 'DATE' | 'TIME' | 'DATETIME' | 'INTERVAL' | 'GEOGRAPHY' | 'NUMERIC' | 'BIGNUMERIC' | 'JSON' | 'ARRAY' | 'STRUCT' | 'RANGE' — Required. The top level type of this field. Can be any GoogleSQL data type (e.g., "INT64", "DATE", "ARRAY").
  - `transformColumns` TransformColumn[] — Output only. This field will be populated if a TRANSFORM clause was used to train a model. TRANSFORM clause (if used) takes feature_columns as input and outputs transform_columns. transform_columns then are used to train the model.
    - `name` string — Output only. Name of the column.
    - `transformSql` string — Output only. The SQL expression used in the column transform.
    - `type` StandardSqlDataType — The data type of a variable such as a function argument. Examples include: * INT64: `{"typeKind": "INT64"}` * ARRAY: { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "STRING"} } * STRUCT>: { "typeKind": "STRUCT", "structType": { "fields": [ { "name": "x", "type": {"typeKind": "STRING"} }, { "name": "y", "type": { "typeKind": "ARRAY", "arrayElementType": {"typeKind": "DATE"} } } ] } } * RANGE: { "typeKind": "RANGE", "rangeElementType": {"typeKind": "DATE"} }
      - `arrayElementType` StandardSqlDataType — recursive
      - `rangeElementType` StandardSqlDataType — recursive
      - `structType` StandardSqlStructType — The representation of a SQL STRUCT type.
        - `fields` StandardSqlField[] — Fields within the struct.
          - `name` string — Optional. The name of this field. Can be absent for struct fields.
          - `type` StandardSqlDataType — recursive
      - `typeKind` 'TYPE_KIND_UNSPECIFIED' | 'INT64' | 'BOOL' | 'FLOAT64' | 'STRING' | 'BYTES' | 'TIMESTAMP' | 'DATE' | 'TIME' | 'DATETIME' | 'INTERVAL' | 'GEOGRAPHY' | 'NUMERIC' | 'BIGNUMERIC' | 'JSON' | 'ARRAY' | 'STRUCT' | 'RANGE' — Required. The top level type of this field. Can be any GoogleSQL data type (e.g., "INT64", "DATE", "ARRAY").
  - `lastModifiedTime` string, int64 — Output only. The time when this model was last modified, in millisecs since the epoch.
  - `modelType` 'MODEL_TYPE_UNSPECIFIED' | 'LINEAR_REGRESSION' | 'LOGISTIC_REGRESSION' | 'KMEANS' | 'MATRIX_FACTORIZATION' | 'DNN_CLASSIFIER' | 'TENSORFLOW' | 'DNN_REGRESSOR' | 'XGBOOST' | 'BOOSTED_TREE_REGRESSOR' | 'BOOSTED_TREE_CLASSIFIER' | 'ARIMA' | 'AUTOML_REGRESSOR' | 'AUTOML_CLASSIFIER' | 'PCA' | 'DNN_LINEAR_COMBINED_CLASSIFIER' | 'DNN_LINEAR_COMBINED_REGRESSOR' | 'AUTOENCODER' | 'ARIMA_PLUS' | 'ARIMA_PLUS_XREG' | 'RANDOM_FOREST_REGRESSOR' | 'RANDOM_FOREST_CLASSIFIER' | 'TENSORFLOW_LITE' | 'ONNX' | 'TRANSFORM_ONLY' | 'CONTRIBUTION_ANALYSIS' — Output only. Type of the model resource.
  - `optimalTrialIds` string[] — Output only. For single-objective [hyperparameter tuning](https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-hp-tuning-overview) models, it only contains the best trial. For multi-objective [hyperparameter tuning](https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-hp-tuning-overview) models, it contains all Pareto optimal trials sorted by trial_id.
  - `remoteModelInfo` RemoteModelInfo — Remote Model Info
    - `maxBatchingRows` string, int64 — Output only. Max number of rows in each batch sent to the remote service. If unset, the number of rows in each batch is set dynamically.
    - `remoteModelVersion` string — Output only. The model version for LLM.
    - `connection` string — Output only. Fully qualified name of the user-provided connection object of the remote model. Format: ```"projects/{project_id}/locations/{location_id}/connections/{connection_id}"```
    - `remoteServiceType` 'REMOTE_SERVICE_TYPE_UNSPECIFIED' | 'CLOUD_AI_TRANSLATE_V3' | 'CLOUD_AI_VISION_V1' | 'CLOUD_AI_NATURAL_LANGUAGE_V1' | 'CLOUD_AI_SPEECH_TO_TEXT_V2' — Output only. The remote service type for remote model.
    - `speechRecognizer` string — Output only. The name of the speech recognizer to use for speech recognition. The expected format is `projects/{project}/locations/{location}/recognizers/{recognizer}`. Customers can specify this field at model creation. If not specified, a default recognizer `projects/{model project}/locations/global/recognizers/_` will be used. See more details at [recognizers](https://cloud.google.com/speech-to-text/v2/docs/reference/rest/v2/projects.locations.recognizers)
    - `endpoint` string — Output only. The endpoint for remote model.
  - `bestTrialId` string, int64 — The best trial_id across all training runs.

## Response `200`

Successful response

---

[API](https://skmtc.dev/google/apis/bigquery.md) · [All operations](https://skmtc.dev/google/apis/bigquery/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/google/bigquery/revisions/98f33ea9c8c9/schema)