Blame - docs/dyn/bigquery_v2.models.html - platform/external/python/google-api-python-client

"evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the

149

# end of training.

150

# data or just the eval data based on whether eval data was used during

151

# training. These are not present for imported models.

152

"clusteringMetrics": { # Evaluation metrics for clustering models. # [Beta] Populated for clustering models.

153

"meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.

154

"daviesBouldinIndex": 3.14, # Davies-Bouldin index.

155

},

156

"regressionMetrics": { # Evaluation metrics for regression models. # Populated for regression models.

157

"meanSquaredLogError": 3.14, # Mean squared log error.

158

"meanAbsoluteError": 3.14, # Mean absolute error.

159

"meanSquaredError": 3.14, # Mean squared error.

160

"medianAbsoluteError": 3.14, # Median absolute error.

161

"rSquared": 3.14, # R^2 score.

162

},

163

"binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.

164

"negativeLabel": "A String", # Label representing the negative class.

165

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

166

# models, the metrics are either macro-averaged or micro-averaged. When

167

# macro-averaged, the metrics are calculated for each label and then an

168

# unweighted average is taken of those values. When micro-averaged, the

169

# metric is calculated globally by counting the total number of correctly

170

# predicted rows.

171

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

172

# positive prediction. For multiclass this is a macro-averaged metric.

173

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

174

# positive actual labels. For multiclass this is a macro-averaged

175

# metric treating each class as a binary classifier.

176

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

177

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

178

# classification models this is the positive class threshold.

179

# For multi-class classfication models this is the confidence

180

# threshold.

181

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

182

# multiclass this is a micro-averaged metric.

183

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

184

# this is a macro-averaged metric.

185

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

186

# metric.

187

},

188

"positiveLabel": "A String", # Label representing the positive class.

189

"binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.

190

{ # Confusion matrix for binary classification models.

191

"truePositives": "A String", # Number of true samples predicted as true.

192

"recall": 3.14, # Aggregate recall.

193

"precision": 3.14, # Aggregate precision.

194

"falseNegatives": "A String", # Number of false samples predicted as false.

195

"trueNegatives": "A String", # Number of true samples predicted as false.

196

"falsePositives": "A String", # Number of false samples predicted as true.

197

"positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.

},

],

},

"multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.

202

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

203

# models, the metrics are either macro-averaged or micro-averaged. When

204

# macro-averaged, the metrics are calculated for each label and then an

205

# unweighted average is taken of those values. When micro-averaged, the

206

# metric is calculated globally by counting the total number of correctly

207

# predicted rows.

208

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

209

# positive prediction. For multiclass this is a macro-averaged metric.

210

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

211

# positive actual labels. For multiclass this is a macro-averaged

212

# metric treating each class as a binary classifier.

213

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

214

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

215

# classification models this is the positive class threshold.

216

# For multi-class classfication models this is the confidence

217

# threshold.

218

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

219

# multiclass this is a micro-averaged metric.

220

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

221

# this is a macro-averaged metric.

222

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

223

# metric.

224

},

225

"confusionMatrixList": [ # Confusion matrix at different thresholds.

226

{ # Confusion matrix for multi-class classification models.

227

"confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the

228

# confusion matrix.

229

"rows": [ # One row per actual label.

230

{ # A single row in the confusion matrix.

231

"entries": [ # Info describing predicted label distribution.

232

{ # A single entry in the confusion matrix.

233

"predictedLabel": "A String", # The predicted label. For confidence_threshold > 0, we will

234

# also add an entry indicating the number of items under the

235

# confidence threshold.

236

"itemCount": "A String", # Number of items being predicted as this label.

237

},

238

],

239

"actualLabel": "A String", # The original label of this row.

},

],

},

],

},

},

"results": [ # Output of each iteration run, results.size() <= max_iterations.

247

{ # Information about a single iteration of the training run.

248

"index": 42, # Index of the iteration, 0 based.

249

"evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.

250

"durationMs": "A String", # Time taken to run the iteration in milliseconds.

251

"learnRate": 3.14, # Learn rate used for this iteration.

252

"trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.

253

"clusterInfos": [ # [Beta] Information about top clusters for clustering models.

254

{ # Information about a single cluster for clustering model.

255

"centroidId": "A String", # Centroid id.

256

"clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.

257

"clusterRadius": 3.14, # Cluster radius, the average distance from centroid

258

# to each point assigned to the cluster.

},

],

},

],

"startTime": "A String", # The start time of this training run.

264

"trainingOptions": { # Options that were used for this training run, includes

265

# user specified and default options that were used.

266

"optimizationStrategy": "A String", # Optimization strategy for training linear regression models.

267

"inputLabelColumns": [ # Name of input label columns in training data.

268

"A String",

269

],

270

"maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative

271

# training algorithms.

272

"earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly

273

# any more (compared to min_relative_progress). Used only for iterative

274

# training algorithms.

275

"initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate

276

# strategy.

277

"dataSplitColumn": "A String", # The column to split data with. This column won't be used as a

278

# feature.

279

# 1. When data_split_method is CUSTOM, the corresponding column should

280

# be boolean. The rows with true value tag are eval data, and the false

281

# are training data.

282

# 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION

283

# rows (from smallest to largest) in the corresponding column are used

284

# as training data, and the rest are eval data. It respects the order

285

# in Orderable data types:

286

# https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties

287

"numClusters": "A String", # [Beta] Number of clusters for clustering models.

288

"l1Regularization": 3.14, # L1 regularization coefficient.

289

"dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.

290

"distanceType": "A String", # [Beta] Distance type for clustering models.

291

"warmStart": True or False, # Whether to train a model from the last checkpoint.

292

"labelClassWeights": { # Weights associated with each label class, for rebalancing the

293

# training data. Only applicable for classification models.

294

"a_key": 3.14,

295

},

296

"lossType": "A String", # Type of loss function used during training run.

297

"dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest

298

# of data will be used as training data. The format should be double.

299

# Accurate to two decimal places.

300

# Default value is 0.2.

301

"l2Regularization": 3.14, # L2 regularization coefficient.

302

"modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only

303

# applicable for imported models.

304

"learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.

305

"minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is

306

# less than 'min_relative_progress'. Used only for iterative training

307

# algorithms.

308

"learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.

},

},

],

"featureColumns": [ # Output only. Input feature columns that were used to train this model.

313

{ # A field or a column.

314

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

315

# specified (e.g., CREATE FUNCTION statement can omit the return type;

316

# in this case the output parameter does not have this "type" field).

317

# Examples:

318

# INT64: {type_kind="INT64"}

319

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

320

# STRUCT<x STRING, y ARRAY<DATE>>:

321

# {type_kind="STRUCT",

322

# struct_type={fields=[

323

# {name="x", type={type_kind="STRING"}},

324

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

325

# ]}}

326

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

327

"fields": [

328

# Object with schema name: StandardSqlField

329

],

330

},

331

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

332

"typeKind": "A String", # Required. The top level type of this field.

333

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

334

},

335

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

336

},

337

],

338

"labels": { # [Optional] The labels associated with this model. You can use these to

339

# organize and group your models. Label keys and values can be no longer

340

# than 63 characters, can only contain lowercase letters, numeric

341

# characters, underscores and dashes. International characters are allowed.

342

# Label values are optional. Label keys must start with a letter and each

343

# label in the list must have a different key.

344

"a_key": "A String",

345

},

346

"creationTime": "A String", # Output only. The time when this model was created, in millisecs since the

347

# epoch.

348

"modelType": "A String", # Output only. Type of the model resource.

349

"modelReference": { # Id path of a model. # Required. Unique identifier for this model.

350

"projectId": "A String", # [Required] The ID of the project containing this model.

351

"datasetId": "A String", # [Required] The ID of the dataset containing this model.

352

"modelId": "A String", # [Required] The ID of the model. The ID must contain only

353

# letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum

354

# length is 1,024 characters.

355

},

356

"etag": "A String", # Output only. A hash of this resource.

357

"location": "A String", # Output only. The geographic location where the model resides. This value

358

# is inherited from the dataset.

359

"friendlyName": "A String", # [Optional] A descriptive name for this model.

360

"expirationTime": "A String", # [Optional] The time when this model expires, in milliseconds since the

361

# epoch. If not present, the model will persist indefinitely. Expired models

362

# will be deleted and their storage reclaimed. The defaultTableExpirationMs

363

# property of the encapsulating dataset can be used to set a default

364

# expirationTime on newly created models.

365

"lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs

# since the epoch.

}</pre>

</div>

<code class="details" id="list">list(projectId, datasetId, pageToken=None, maxResults=None)</code>

372

<pre>Lists all models in the specified dataset. Requires the READER dataset

role.

Args:

projectId: string, Project ID of the models to list. (required)

377

datasetId: string, Dataset ID of the models to list. (required)

378

pageToken: string, Page token, returned by a previous call to request the next page of

379

results

380

maxResults: integer, The maximum number of results per page.

381

382

Returns:

383

An object of the form:

384

385

{

386

"models": [ # Models in the requested dataset. Only the following fields are populated:

387

# model_reference, model_type, creation_time, last_modified_time and

388

# labels.

389

{

390

"labelColumns": [ # Output only. Label columns that were used to train this model.

391

# The output of the model will have a "predicted_" prefix to these columns.

392

{ # A field or a column.

393

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

394

# specified (e.g., CREATE FUNCTION statement can omit the return type;

395

# in this case the output parameter does not have this "type" field).

396

# Examples:

397

# INT64: {type_kind="INT64"}

398

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

399

# STRUCT<x STRING, y ARRAY<DATE>>:

400

# {type_kind="STRUCT",

401

# struct_type={fields=[

402

# {name="x", type={type_kind="STRING"}},

403

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

404

# ]}}

405

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

406

"fields": [

407

# Object with schema name: StandardSqlField

408

],

409

},

410

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

411

"typeKind": "A String", # Required. The top level type of this field.

412

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

413

},

414

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

415

},

416

],

417

"description": "A String", # [Optional] A user-friendly description of this model.

418

"trainingRuns": [ # Output only. Information for all training runs in increasing order of

419

# start_time.

420

{ # Information about a single training query run for the model.

421

"evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the

422

# end of training.

423

# data or just the eval data based on whether eval data was used during

424

# training. These are not present for imported models.

425

"clusteringMetrics": { # Evaluation metrics for clustering models. # [Beta] Populated for clustering models.

426

"meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.

427

"daviesBouldinIndex": 3.14, # Davies-Bouldin index.

428

},

429

"regressionMetrics": { # Evaluation metrics for regression models. # Populated for regression models.

430

"meanSquaredLogError": 3.14, # Mean squared log error.

431

"meanAbsoluteError": 3.14, # Mean absolute error.

432

"meanSquaredError": 3.14, # Mean squared error.

433

"medianAbsoluteError": 3.14, # Median absolute error.

434

"rSquared": 3.14, # R^2 score.

435

},

436

"binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.

437

"negativeLabel": "A String", # Label representing the negative class.

438

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

439

# models, the metrics are either macro-averaged or micro-averaged. When

440

# macro-averaged, the metrics are calculated for each label and then an

441

# unweighted average is taken of those values. When micro-averaged, the

442

# metric is calculated globally by counting the total number of correctly

443

# predicted rows.

444

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

445

# positive prediction. For multiclass this is a macro-averaged metric.

446

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

447

# positive actual labels. For multiclass this is a macro-averaged

448

# metric treating each class as a binary classifier.

449

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

450

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

451

# classification models this is the positive class threshold.

452

# For multi-class classfication models this is the confidence

453

# threshold.

454

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

455

# multiclass this is a micro-averaged metric.

456

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

457

# this is a macro-averaged metric.

458

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

459

# metric.

460

},

461

"positiveLabel": "A String", # Label representing the positive class.

462

"binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.

463

{ # Confusion matrix for binary classification models.

464

"truePositives": "A String", # Number of true samples predicted as true.

465

"recall": 3.14, # Aggregate recall.

466

"precision": 3.14, # Aggregate precision.

467

"falseNegatives": "A String", # Number of false samples predicted as false.

468

"trueNegatives": "A String", # Number of true samples predicted as false.

469

"falsePositives": "A String", # Number of false samples predicted as true.

470

"positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.

},

],

},

"multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.

475

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

476

# models, the metrics are either macro-averaged or micro-averaged. When

477

# macro-averaged, the metrics are calculated for each label and then an

478

# unweighted average is taken of those values. When micro-averaged, the

479

# metric is calculated globally by counting the total number of correctly

480

# predicted rows.

481

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

482

# positive prediction. For multiclass this is a macro-averaged metric.

483

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

484

# positive actual labels. For multiclass this is a macro-averaged

485

# metric treating each class as a binary classifier.

486

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

487

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

488

# classification models this is the positive class threshold.

489

# For multi-class classfication models this is the confidence

490

# threshold.

491

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

492

# multiclass this is a micro-averaged metric.

493

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

494

# this is a macro-averaged metric.

495

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

496

# metric.

497

},

498

"confusionMatrixList": [ # Confusion matrix at different thresholds.

499

{ # Confusion matrix for multi-class classification models.

500

"confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the

501

# confusion matrix.

502

"rows": [ # One row per actual label.

503

{ # A single row in the confusion matrix.

504

"entries": [ # Info describing predicted label distribution.

505

{ # A single entry in the confusion matrix.

506

"predictedLabel": "A String", # The predicted label. For confidence_threshold > 0, we will

507

# also add an entry indicating the number of items under the

508

# confidence threshold.

509

"itemCount": "A String", # Number of items being predicted as this label.

510

},

511

],

512

"actualLabel": "A String", # The original label of this row.

},

],

},

],

},

},

"results": [ # Output of each iteration run, results.size() <= max_iterations.

520

{ # Information about a single iteration of the training run.

521

"index": 42, # Index of the iteration, 0 based.

522

"evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.

523

"durationMs": "A String", # Time taken to run the iteration in milliseconds.

524

"learnRate": 3.14, # Learn rate used for this iteration.

525

"trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.

526

"clusterInfos": [ # [Beta] Information about top clusters for clustering models.

527

{ # Information about a single cluster for clustering model.

528

"centroidId": "A String", # Centroid id.

529

"clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.

530

"clusterRadius": 3.14, # Cluster radius, the average distance from centroid

531

# to each point assigned to the cluster.

},

],

},

],

"startTime": "A String", # The start time of this training run.

537

"trainingOptions": { # Options that were used for this training run, includes

538

# user specified and default options that were used.

539

"optimizationStrategy": "A String", # Optimization strategy for training linear regression models.

540

"inputLabelColumns": [ # Name of input label columns in training data.

541

"A String",

542

],

543

"maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative

544

# training algorithms.

545

"earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly

546

# any more (compared to min_relative_progress). Used only for iterative

547

# training algorithms.

548

"initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate

549

# strategy.

550

"dataSplitColumn": "A String", # The column to split data with. This column won't be used as a

551

# feature.

552

# 1. When data_split_method is CUSTOM, the corresponding column should

553

# be boolean. The rows with true value tag are eval data, and the false

554

# are training data.

555

# 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION

556

# rows (from smallest to largest) in the corresponding column are used

557

# as training data, and the rest are eval data. It respects the order

558

# in Orderable data types:

559

# https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties

560

"numClusters": "A String", # [Beta] Number of clusters for clustering models.

561

"l1Regularization": 3.14, # L1 regularization coefficient.

562

"dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.

563

"distanceType": "A String", # [Beta] Distance type for clustering models.

564

"warmStart": True or False, # Whether to train a model from the last checkpoint.

565

"labelClassWeights": { # Weights associated with each label class, for rebalancing the

566

# training data. Only applicable for classification models.

567

"a_key": 3.14,

568

},

569

"lossType": "A String", # Type of loss function used during training run.

570

"dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest

571

# of data will be used as training data. The format should be double.

572

# Accurate to two decimal places.

573

# Default value is 0.2.

574

"l2Regularization": 3.14, # L2 regularization coefficient.

575

"modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only

576

# applicable for imported models.

577

"learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.

578

"minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is

579

# less than 'min_relative_progress'. Used only for iterative training

580

# algorithms.

581

"learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.

},

},

],

"featureColumns": [ # Output only. Input feature columns that were used to train this model.

586

{ # A field or a column.

587

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

588

# specified (e.g., CREATE FUNCTION statement can omit the return type;

589

# in this case the output parameter does not have this "type" field).

590

# Examples:

591

# INT64: {type_kind="INT64"}

592

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

593

# STRUCT<x STRING, y ARRAY<DATE>>:

594

# {type_kind="STRUCT",

595

# struct_type={fields=[

596

# {name="x", type={type_kind="STRING"}},

597

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

598

# ]}}

599

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

600

"fields": [

601

# Object with schema name: StandardSqlField

602

],

603

},

604

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

605

"typeKind": "A String", # Required. The top level type of this field.

606

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

607

},

608

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

609

},

610

],

611

"labels": { # [Optional] The labels associated with this model. You can use these to

612

# organize and group your models. Label keys and values can be no longer

613

# than 63 characters, can only contain lowercase letters, numeric

614

# characters, underscores and dashes. International characters are allowed.

615

# Label values are optional. Label keys must start with a letter and each

616

# label in the list must have a different key.

617

"a_key": "A String",

618

},

619

"creationTime": "A String", # Output only. The time when this model was created, in millisecs since the

620

# epoch.

621

"modelType": "A String", # Output only. Type of the model resource.

622

"modelReference": { # Id path of a model. # Required. Unique identifier for this model.

623

"projectId": "A String", # [Required] The ID of the project containing this model.

624

"datasetId": "A String", # [Required] The ID of the dataset containing this model.

625

"modelId": "A String", # [Required] The ID of the model. The ID must contain only

626

# letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum

627

# length is 1,024 characters.

628

},

629

"etag": "A String", # Output only. A hash of this resource.

630

"location": "A String", # Output only. The geographic location where the model resides. This value

631

# is inherited from the dataset.

632

"friendlyName": "A String", # [Optional] A descriptive name for this model.

633

"expirationTime": "A String", # [Optional] The time when this model expires, in milliseconds since the

634

# epoch. If not present, the model will persist indefinitely. Expired models

635

# will be deleted and their storage reclaimed. The defaultTableExpirationMs

636

# property of the encapsulating dataset can be used to set a default

637

# expirationTime on newly created models.

638

"lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs

# since the epoch.

},

],

"nextPageToken": "A String", # A token to request the next page of results.

}</pre>

</div>

<code class="details" id="list_next">list_next(previous_request, previous_response)</code>

648

<pre>Retrieves the next page of results.

649

650

Args:

651

previous_request: The request for the previous page. (required)

652

previous_response: The response from the request for the previous page. (required)

653

654

Returns:

655

A request object that you can call 'execute()' on to request the next

656

page. Returns None if there are no more items in the collection.

</pre>

</div>

<code class="details" id="patch">patch(projectId, datasetId, modelId, body)</code>

662

<pre>Patch specific fields in the specified model.

663

664

Args:

665

projectId: string, Project ID of the model to patch. (required)

666

datasetId: string, Dataset ID of the model to patch. (required)

667

modelId: string, Model ID of the model to patch. (required)

668

body: object, The request body. (required)

669

The object takes the form of:

670

671

{

672

"labelColumns": [ # Output only. Label columns that were used to train this model.

673

# The output of the model will have a "predicted_" prefix to these columns.

674

{ # A field or a column.

675

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

676

# specified (e.g., CREATE FUNCTION statement can omit the return type;

677

# in this case the output parameter does not have this "type" field).

678

# Examples:

679

# INT64: {type_kind="INT64"}

680

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

681

# STRUCT<x STRING, y ARRAY<DATE>>:

682

# {type_kind="STRUCT",

683

# struct_type={fields=[

684

# {name="x", type={type_kind="STRING"}},

685

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

686

# ]}}

687

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

688

"fields": [

689

# Object with schema name: StandardSqlField

690

],

691

},

692

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

693

"typeKind": "A String", # Required. The top level type of this field.

694

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

695

},

696

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

697

},

698

],

699

"description": "A String", # [Optional] A user-friendly description of this model.

700

"trainingRuns": [ # Output only. Information for all training runs in increasing order of

701

# start_time.

702

{ # Information about a single training query run for the model.

703

"evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the

704

# end of training.

705

# data or just the eval data based on whether eval data was used during

706

# training. These are not present for imported models.

707

"clusteringMetrics": { # Evaluation metrics for clustering models. # [Beta] Populated for clustering models.

708

"meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.

709

"daviesBouldinIndex": 3.14, # Davies-Bouldin index.

710

},

711

"regressionMetrics": { # Evaluation metrics for regression models. # Populated for regression models.

712

"meanSquaredLogError": 3.14, # Mean squared log error.

713

"meanAbsoluteError": 3.14, # Mean absolute error.

714

"meanSquaredError": 3.14, # Mean squared error.

715

"medianAbsoluteError": 3.14, # Median absolute error.

716

"rSquared": 3.14, # R^2 score.

717

},

718

"binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.

719

"negativeLabel": "A String", # Label representing the negative class.

720

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

721

# models, the metrics are either macro-averaged or micro-averaged. When

722

# macro-averaged, the metrics are calculated for each label and then an

723

# unweighted average is taken of those values. When micro-averaged, the

724

# metric is calculated globally by counting the total number of correctly

725

# predicted rows.

726

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

727

# positive prediction. For multiclass this is a macro-averaged metric.

728

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

729

# positive actual labels. For multiclass this is a macro-averaged

730

# metric treating each class as a binary classifier.

731

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

732

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

733

# classification models this is the positive class threshold.

734

# For multi-class classfication models this is the confidence

735

# threshold.

736

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

737

# multiclass this is a micro-averaged metric.

738

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

739

# this is a macro-averaged metric.

740

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

741

# metric.

742

},

743

"positiveLabel": "A String", # Label representing the positive class.

744

"binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.

745

{ # Confusion matrix for binary classification models.

746

"truePositives": "A String", # Number of true samples predicted as true.

747

"recall": 3.14, # Aggregate recall.

748

"precision": 3.14, # Aggregate precision.

749

"falseNegatives": "A String", # Number of false samples predicted as false.

750

"trueNegatives": "A String", # Number of true samples predicted as false.

751

"falsePositives": "A String", # Number of false samples predicted as true.

752

"positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.

},

],

},

"multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.

757

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

758

# models, the metrics are either macro-averaged or micro-averaged. When

759

# macro-averaged, the metrics are calculated for each label and then an

760

# unweighted average is taken of those values. When micro-averaged, the

761

# metric is calculated globally by counting the total number of correctly

762

# predicted rows.

763

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

764

# positive prediction. For multiclass this is a macro-averaged metric.

765

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

766

# positive actual labels. For multiclass this is a macro-averaged

767

# metric treating each class as a binary classifier.

768

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

769

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

770

# classification models this is the positive class threshold.

771

# For multi-class classfication models this is the confidence

772

# threshold.

773

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

774

# multiclass this is a micro-averaged metric.

775

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

776

# this is a macro-averaged metric.

777

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

778

# metric.

779

},

780

"confusionMatrixList": [ # Confusion matrix at different thresholds.

781

{ # Confusion matrix for multi-class classification models.

782

"confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the

783

# confusion matrix.

784

"rows": [ # One row per actual label.

785

{ # A single row in the confusion matrix.

786

"entries": [ # Info describing predicted label distribution.

787

{ # A single entry in the confusion matrix.

788

"predictedLabel": "A String", # The predicted label. For confidence_threshold > 0, we will

789

# also add an entry indicating the number of items under the

790

# confidence threshold.

791

"itemCount": "A String", # Number of items being predicted as this label.

792

},

793

],

794

"actualLabel": "A String", # The original label of this row.

},

],

},

],

},

},

"results": [ # Output of each iteration run, results.size() <= max_iterations.

802

{ # Information about a single iteration of the training run.

803

"index": 42, # Index of the iteration, 0 based.

804

"evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.

805

"durationMs": "A String", # Time taken to run the iteration in milliseconds.

806

"learnRate": 3.14, # Learn rate used for this iteration.

807

"trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.

808

"clusterInfos": [ # [Beta] Information about top clusters for clustering models.

809

{ # Information about a single cluster for clustering model.

810

"centroidId": "A String", # Centroid id.

811

"clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.

812

"clusterRadius": 3.14, # Cluster radius, the average distance from centroid

813

# to each point assigned to the cluster.

},

],

},

],

"startTime": "A String", # The start time of this training run.

819

"trainingOptions": { # Options that were used for this training run, includes

820

# user specified and default options that were used.

821

"optimizationStrategy": "A String", # Optimization strategy for training linear regression models.

822

"inputLabelColumns": [ # Name of input label columns in training data.

823

"A String",

824

],

825

"maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative

826

# training algorithms.

827

"earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly

828

# any more (compared to min_relative_progress). Used only for iterative

829

# training algorithms.

830

"initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate

831

# strategy.

832

"dataSplitColumn": "A String", # The column to split data with. This column won't be used as a

833

# feature.

834

# 1. When data_split_method is CUSTOM, the corresponding column should

835

# be boolean. The rows with true value tag are eval data, and the false

836

# are training data.

837

# 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION

838

# rows (from smallest to largest) in the corresponding column are used

839

# as training data, and the rest are eval data. It respects the order

840

# in Orderable data types:

841

# https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties

842

"numClusters": "A String", # [Beta] Number of clusters for clustering models.

843

"l1Regularization": 3.14, # L1 regularization coefficient.

844

"dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.

845

"distanceType": "A String", # [Beta] Distance type for clustering models.

846

"warmStart": True or False, # Whether to train a model from the last checkpoint.

847

"labelClassWeights": { # Weights associated with each label class, for rebalancing the

848

# training data. Only applicable for classification models.

849

"a_key": 3.14,

850

},

851

"lossType": "A String", # Type of loss function used during training run.

852

"dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest

853

# of data will be used as training data. The format should be double.

854

# Accurate to two decimal places.

855

# Default value is 0.2.

856

"l2Regularization": 3.14, # L2 regularization coefficient.

857

"modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only

858

# applicable for imported models.

859

"learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.

860

"minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is

861

# less than 'min_relative_progress'. Used only for iterative training

862

# algorithms.

863

"learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.

},

},

],

"featureColumns": [ # Output only. Input feature columns that were used to train this model.

868

{ # A field or a column.

869

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

870

# specified (e.g., CREATE FUNCTION statement can omit the return type;

871

# in this case the output parameter does not have this "type" field).

872

# Examples:

873

# INT64: {type_kind="INT64"}

874

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

875

# STRUCT<x STRING, y ARRAY<DATE>>:

876

# {type_kind="STRUCT",

877

# struct_type={fields=[

878

# {name="x", type={type_kind="STRING"}},

879

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

880

# ]}}

881

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

882

"fields": [

883

# Object with schema name: StandardSqlField

884

],

885

},

886

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

887

"typeKind": "A String", # Required. The top level type of this field.

888

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

889

},

890

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

891

},

892

],

893

"labels": { # [Optional] The labels associated with this model. You can use these to

894

# organize and group your models. Label keys and values can be no longer

895

# than 63 characters, can only contain lowercase letters, numeric

896

# characters, underscores and dashes. International characters are allowed.

897

# Label values are optional. Label keys must start with a letter and each

898

# label in the list must have a different key.

899

"a_key": "A String",

900

},

901

"creationTime": "A String", # Output only. The time when this model was created, in millisecs since the

902

# epoch.

903

"modelType": "A String", # Output only. Type of the model resource.

904

"modelReference": { # Id path of a model. # Required. Unique identifier for this model.

905

"projectId": "A String", # [Required] The ID of the project containing this model.

906

"datasetId": "A String", # [Required] The ID of the dataset containing this model.

907

"modelId": "A String", # [Required] The ID of the model. The ID must contain only

908

# letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum

909

# length is 1,024 characters.

910

},

911

"etag": "A String", # Output only. A hash of this resource.

912

"location": "A String", # Output only. The geographic location where the model resides. This value

913

# is inherited from the dataset.

914

"friendlyName": "A String", # [Optional] A descriptive name for this model.

915

"expirationTime": "A String", # [Optional] The time when this model expires, in milliseconds since the

916

# epoch. If not present, the model will persist indefinitely. Expired models

917

# will be deleted and their storage reclaimed. The defaultTableExpirationMs

918

# property of the encapsulating dataset can be used to set a default

919

# expirationTime on newly created models.

920

"lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs

# since the epoch.

}

Returns:

An object of the form:

927

928

{

929

"labelColumns": [ # Output only. Label columns that were used to train this model.

930

# The output of the model will have a "predicted_" prefix to these columns.

931

{ # A field or a column.

932

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

933

# specified (e.g., CREATE FUNCTION statement can omit the return type;

934

# in this case the output parameter does not have this "type" field).

935

# Examples:

936

# INT64: {type_kind="INT64"}

937

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

938

# STRUCT<x STRING, y ARRAY<DATE>>:

939

# {type_kind="STRUCT",

940

# struct_type={fields=[

941

# {name="x", type={type_kind="STRING"}},

942

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

943

# ]}}

944

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

945

"fields": [

946

# Object with schema name: StandardSqlField

947

],

948

},

949

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

950

"typeKind": "A String", # Required. The top level type of this field.

951

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

952

},

953

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

954

},

955

],

956

"description": "A String", # [Optional] A user-friendly description of this model.

957

"trainingRuns": [ # Output only. Information for all training runs in increasing order of

958

# start_time.

959

{ # Information about a single training query run for the model.

960

"evaluationMetrics": { # Evaluation metrics of a model. These are either computed on all training # The evaluation metrics over training/eval data that were computed at the

961

# end of training.

962

# data or just the eval data based on whether eval data was used during

963

# training. These are not present for imported models.

964

"clusteringMetrics": { # Evaluation metrics for clustering models. # [Beta] Populated for clustering models.

965

"meanSquaredDistance": 3.14, # Mean of squared distances between each sample to its cluster centroid.

966

"daviesBouldinIndex": 3.14, # Davies-Bouldin index.

967

},

968

"regressionMetrics": { # Evaluation metrics for regression models. # Populated for regression models.

969

"meanSquaredLogError": 3.14, # Mean squared log error.

970

"meanAbsoluteError": 3.14, # Mean absolute error.

971

"meanSquaredError": 3.14, # Mean squared error.

972

"medianAbsoluteError": 3.14, # Median absolute error.

973

"rSquared": 3.14, # R^2 score.

974

},

975

"binaryClassificationMetrics": { # Evaluation metrics for binary classification/classifier models. # Populated for binary classification/classifier models.

976

"negativeLabel": "A String", # Label representing the negative class.

977

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

978

# models, the metrics are either macro-averaged or micro-averaged. When

979

# macro-averaged, the metrics are calculated for each label and then an

980

# unweighted average is taken of those values. When micro-averaged, the

981

# metric is calculated globally by counting the total number of correctly

982

# predicted rows.

983

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

984

# positive prediction. For multiclass this is a macro-averaged metric.

985

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

986

# positive actual labels. For multiclass this is a macro-averaged

987

# metric treating each class as a binary classifier.

988

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

989

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

990

# classification models this is the positive class threshold.

991

# For multi-class classfication models this is the confidence

992

# threshold.

993

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

994

# multiclass this is a micro-averaged metric.

995

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

996

# this is a macro-averaged metric.

997

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

998

# metric.

999

},

1000

"positiveLabel": "A String", # Label representing the positive class.

1001

"binaryConfusionMatrixList": [ # Binary confusion matrix at multiple thresholds.

1002

{ # Confusion matrix for binary classification models.

1003

"truePositives": "A String", # Number of true samples predicted as true.

1004

"recall": 3.14, # Aggregate recall.

1005

"precision": 3.14, # Aggregate precision.

1006

"falseNegatives": "A String", # Number of false samples predicted as false.

1007

"trueNegatives": "A String", # Number of true samples predicted as false.

1008

"falsePositives": "A String", # Number of false samples predicted as true.

1009

"positiveClassThreshold": 3.14, # Threshold value used when computing each of the following metric.

},

],

},

"multiClassClassificationMetrics": { # Evaluation metrics for multi-class classification/classifier models. # Populated for multi-class classification/classifier models.

1014

"aggregateClassificationMetrics": { # Aggregate metrics for classification/classifier models. For multi-class # Aggregate classification metrics.

1015

# models, the metrics are either macro-averaged or micro-averaged. When

1016

# macro-averaged, the metrics are calculated for each label and then an

1017

# unweighted average is taken of those values. When micro-averaged, the

1018

# metric is calculated globally by counting the total number of correctly

1019

# predicted rows.

1020

"recall": 3.14, # Recall is the fraction of actual positive labels that were given a

1021

# positive prediction. For multiclass this is a macro-averaged metric.

1022

"precision": 3.14, # Precision is the fraction of actual positive predictions that had

1023

# positive actual labels. For multiclass this is a macro-averaged

1024

# metric treating each class as a binary classifier.

1025

"logLoss": 3.14, # Logarithmic Loss. For multiclass this is a macro-averaged metric.

1026

"threshold": 3.14, # Threshold at which the metrics are computed. For binary

1027

# classification models this is the positive class threshold.

1028

# For multi-class classfication models this is the confidence

1029

# threshold.

1030

"accuracy": 3.14, # Accuracy is the fraction of predictions given the correct label. For

1031

# multiclass this is a micro-averaged metric.

1032

"f1Score": 3.14, # The F1 score is an average of recall and precision. For multiclass

1033

# this is a macro-averaged metric.

1034

"rocAuc": 3.14, # Area Under a ROC Curve. For multiclass this is a macro-averaged

1035

# metric.

1036

},

1037

"confusionMatrixList": [ # Confusion matrix at different thresholds.

1038

{ # Confusion matrix for multi-class classification models.

1039

"confidenceThreshold": 3.14, # Confidence threshold used when computing the entries of the

1040

# confusion matrix.

1041

"rows": [ # One row per actual label.

1042

{ # A single row in the confusion matrix.

1043

"entries": [ # Info describing predicted label distribution.

1044

{ # A single entry in the confusion matrix.

1045

"predictedLabel": "A String", # The predicted label. For confidence_threshold > 0, we will

1046

# also add an entry indicating the number of items under the

1047

# confidence threshold.

1048

"itemCount": "A String", # Number of items being predicted as this label.

1049

},

1050

],

1051

"actualLabel": "A String", # The original label of this row.

},

],

},

],

},

},

"results": [ # Output of each iteration run, results.size() <= max_iterations.

1059

{ # Information about a single iteration of the training run.

1060

"index": 42, # Index of the iteration, 0 based.

1061

"evalLoss": 3.14, # Loss computed on the eval data at the end of iteration.

1062

"durationMs": "A String", # Time taken to run the iteration in milliseconds.

1063

"learnRate": 3.14, # Learn rate used for this iteration.

1064

"trainingLoss": 3.14, # Loss computed on the training data at the end of iteration.

1065

"clusterInfos": [ # [Beta] Information about top clusters for clustering models.

1066

{ # Information about a single cluster for clustering model.

1067

"centroidId": "A String", # Centroid id.

1068

"clusterSize": "A String", # Cluster size, the total number of points assigned to the cluster.

1069

"clusterRadius": 3.14, # Cluster radius, the average distance from centroid

1070

# to each point assigned to the cluster.

},

],

},

],

"startTime": "A String", # The start time of this training run.

1076

"trainingOptions": { # Options that were used for this training run, includes

1077

# user specified and default options that were used.

1078

"optimizationStrategy": "A String", # Optimization strategy for training linear regression models.

1079

"inputLabelColumns": [ # Name of input label columns in training data.

1080

"A String",

1081

],

1082

"maxIterations": "A String", # The maximum number of iterations in training. Used only for iterative

1083

# training algorithms.

1084

"earlyStop": True or False, # Whether to stop early when the loss doesn't improve significantly

1085

# any more (compared to min_relative_progress). Used only for iterative

1086

# training algorithms.

1087

"initialLearnRate": 3.14, # Specifies the initial learning rate for the line search learn rate

1088

# strategy.

1089

"dataSplitColumn": "A String", # The column to split data with. This column won't be used as a

1090

# feature.

1091

# 1. When data_split_method is CUSTOM, the corresponding column should

1092

# be boolean. The rows with true value tag are eval data, and the false

1093

# are training data.

1094

# 2. When data_split_method is SEQ, the first DATA_SPLIT_EVAL_FRACTION

1095

# rows (from smallest to largest) in the corresponding column are used

1096

# as training data, and the rest are eval data. It respects the order

1097

# in Orderable data types:

1098

# https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types#data-type-properties

1099

"numClusters": "A String", # [Beta] Number of clusters for clustering models.

1100

"l1Regularization": 3.14, # L1 regularization coefficient.

1101

"dataSplitMethod": "A String", # The data split type for training and evaluation, e.g. RANDOM.

1102

"distanceType": "A String", # [Beta] Distance type for clustering models.

1103

"warmStart": True or False, # Whether to train a model from the last checkpoint.

1104

"labelClassWeights": { # Weights associated with each label class, for rebalancing the

1105

# training data. Only applicable for classification models.

1106

"a_key": 3.14,

1107

},

1108

"lossType": "A String", # Type of loss function used during training run.

1109

"dataSplitEvalFraction": 3.14, # The fraction of evaluation data over the whole input data. The rest

1110

# of data will be used as training data. The format should be double.

1111

# Accurate to two decimal places.

1112

# Default value is 0.2.

1113

"l2Regularization": 3.14, # L2 regularization coefficient.

1114

"modelUri": "A String", # [Beta] Google Cloud Storage URI from which the model was imported. Only

1115

# applicable for imported models.

1116

"learnRateStrategy": "A String", # The strategy to determine learn rate for the current iteration.

1117

"minRelativeProgress": 3.14, # When early_stop is true, stops training when accuracy improvement is

1118

# less than 'min_relative_progress'. Used only for iterative training

1119

# algorithms.

1120

"learnRate": 3.14, # Learning rate in training. Used only for iterative training algorithms.

},

},

],

"featureColumns": [ # Output only. Input feature columns that were used to train this model.

1125

{ # A field or a column.

1126

"type": { # The type of a variable, e.g., a function argument. # Optional. The type of this parameter. Absent if not explicitly

1127

# specified (e.g., CREATE FUNCTION statement can omit the return type;

1128

# in this case the output parameter does not have this "type" field).

1129

# Examples:

1130

# INT64: {type_kind="INT64"}

1131

# ARRAY<STRING>: {type_kind="ARRAY", array_element_type="STRING"}

1132

# STRUCT<x STRING, y ARRAY<DATE>>:

1133

# {type_kind="STRUCT",

1134

# struct_type={fields=[

1135

# {name="x", type={type_kind="STRING"}},

1136

# {name="y", type={type_kind="ARRAY", array_element_type="DATE"}}

1137

# ]}}

1138

"structType": { # The fields of this struct, in order, if type_kind = "STRUCT".

1139

"fields": [

1140

# Object with schema name: StandardSqlField

1141

],

1142

},

1143

"arrayElementType": # Object with schema name: StandardSqlDataType # The type of the array's elements, if type_kind = "ARRAY".

1144

"typeKind": "A String", # Required. The top level type of this field.

1145

# Can be any standard SQL data type (e.g., "INT64", "DATE", "ARRAY").

1146

},

1147

"name": "A String", # Optional. The name of this field. Can be absent for struct fields.

1148

},

1149

],

1150

"labels": { # [Optional] The labels associated with this model. You can use these to

1151

# organize and group your models. Label keys and values can be no longer

1152

# than 63 characters, can only contain lowercase letters, numeric

1153

# characters, underscores and dashes. International characters are allowed.

1154

# Label values are optional. Label keys must start with a letter and each

1155

# label in the list must have a different key.

1156

"a_key": "A String",

1157

},

1158

"creationTime": "A String", # Output only. The time when this model was created, in millisecs since the

1159

# epoch.

1160

"modelType": "A String", # Output only. Type of the model resource.

1161

"modelReference": { # Id path of a model. # Required. Unique identifier for this model.

1162

"projectId": "A String", # [Required] The ID of the project containing this model.

1163

"datasetId": "A String", # [Required] The ID of the dataset containing this model.

1164

"modelId": "A String", # [Required] The ID of the model. The ID must contain only

1165

# letters (a-z, A-Z), numbers (0-9), or underscores (_). The maximum

1166

# length is 1,024 characters.

1167

},

1168

"etag": "A String", # Output only. A hash of this resource.

1169

"location": "A String", # Output only. The geographic location where the model resides. This value

1170

# is inherited from the dataset.

1171

"friendlyName": "A String", # [Optional] A descriptive name for this model.

1172

"expirationTime": "A String", # [Optional] The time when this model expires, in milliseconds since the

1173

# epoch. If not present, the model will persist indefinitely. Expired models

1174

# will be deleted and their storage reclaimed. The defaultTableExpirationMs

1175

# property of the encapsulating dataset can be used to set a default

1176

# expirationTime on newly created models.

1177

"lastModifiedTime": "A String", # Output only. The time when this model was last modified, in millisecs

# since the epoch.

}</pre>

</div>

</body></html>