Blame - docs/dyn/ml_v1.projects.models.html - platform/external/python/google-api-python-client

2017-03-13 12:12:03 -0400

[diff] [blame]

76

<h2>Instance Methods</h2>

77

78

<code><a href="ml_v1.projects.models.versions.html">versions()</a></code>

79

</p>

80

<p class="firstline">Returns the versions Resource.</p>

81

82

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

83

<code><a href="#create">create(parent, body, x__xgafv=None)</a></code></p>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

84

<p class="firstline">Creates a model which will later contain one or more versions.</p>

85

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

86

<code><a href="#delete">delete(name, x__xgafv=None)</a></code></p>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

87

<p class="firstline">Deletes a model.</p>

88

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

89

<code><a href="#get">get(name, x__xgafv=None)</a></code></p>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

90

<p class="firstline">Gets information about a model, including its name, the description (if</p>

91

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

92

<code><a href="#getIamPolicy">getIamPolicy(resource, x__xgafv=None)</a></code></p>

93

<p class="firstline">Gets the access control policy for a resource.</p>

94

95

<code><a href="#list">list(parent, pageToken=None, x__xgafv=None, pageSize=None, filter=None)</a></code></p>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

96

<p class="firstline">Lists the models in a project.</p>

97

98

<code><a href="#list_next">list_next(previous_request, previous_response)</a></code></p>

99

<p class="firstline">Retrieves the next page of results.</p>

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

100

101

<code><a href="#patch">patch(name, body, updateMask=None, x__xgafv=None)</a></code></p>

102

<p class="firstline">Updates a specific model resource.</p>

103

104

<code><a href="#setIamPolicy">setIamPolicy(resource, body, x__xgafv=None)</a></code></p>

105

<p class="firstline">Sets the access control policy on the specified resource. Replaces any</p>

106

107

<code><a href="#testIamPermissions">testIamPermissions(resource, body, x__xgafv=None)</a></code></p>

108

<p class="firstline">Returns permissions that a caller has on the specified resource.</p>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

109

<h3>Method Details</h3>

110

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

111

<code class="details" id="create">create(parent, body, x__xgafv=None)</code>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

112

<pre>Creates a model which will later contain one or more versions.

113

114

You must add at least one version before you can request predictions from

115

the model. Add versions by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

116

[projects.models.versions.create](/ml-engine/reference/rest/v1/projects.models.versions/create).

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

117

118

Args:

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

119

parent: string, Required. The project name. (required)

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

120

body: object, The request body. (required)

121

The object takes the form of:

122

123

{ # Represents a machine learning solution.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

124

#

125

# A model can have multiple versions, each of which is a deployed, trained

126

# model ready to receive prediction requests. The model itself is just a

127

# container.

128

"description": "A String", # Optional. The description specified for the model when it was created.

129

"onlinePredictionConsoleLogging": True or False, # Optional. If true, online prediction nodes send `stderr` and `stdout`

130

# streams to Stackdriver Logging. These can be more verbose than the standard

131

# access logs (see `onlinePredictionLogging`) and can incur higher cost.

132

# However, they are helpful for debugging. Note that

133

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

134

# your project receives prediction requests at a high QPS. Estimate your

135

# costs before enabling this option.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

136

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

137

# Default is false.

138

"labels": { # Optional. One or more labels that you can add, to organize your models.

139

# Each label is a key-value pair, where both the key and the value are

140

# arbitrary strings that you supply.

141

# For more information, see the documentation on

142

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

143

"a_key": "A String",

144

},

145

"regions": [ # Optional. The list of regions where the model is going to be deployed.

146

# Currently only one region per model is supported.

147

# Defaults to 'us-central1' if nothing is set.

148

# See the <a href="/ml-engine/docs/tensorflow/regions">available regions</a>

149

# for AI Platform services.

150

# Note:

151

# * No matter where a model is deployed, it can always be accessed by

152

# users from anywhere, both for online and batch prediction.

153

# * The region for a batch prediction job is set by the region field when

154

# submitting the batch prediction job and does not take its value from

# this field.

"A String",

],

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

159

# prevent simultaneous updates of a model from overwriting each other.

160

# It is strongly suggested that systems make use of the `etag` in the

161

# read-modify-write cycle to perform model updates in order to avoid race

162

# conditions: An `etag` is returned in the response to `GetModel`, and

163

# systems are expected to put that etag in the request to `UpdateModel` to

164

# ensure that their change will be applied to the model as intended.

165

"defaultVersion": { # Represents a version of the model. # Output only. The default version of the model. This version will be used to

166

# handle prediction requests that do not specify a version.

167

#

168

# You can change the default version by calling

169

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

170

#

171

# Each version is a trained model deployed in the cloud, ready to handle

172

# prediction requests. A model can have multiple versions. You can get

173

# information about all of the versions of a given model by calling

174

# [projects.models.versions.list](/ml-engine/reference/rest/v1/projects.models.versions/list).

175

"errorMessage": "A String", # Output only. The details of a failure or a cancellation.

176

"labels": { # Optional. One or more labels that you can add, to organize your model

177

# versions. Each label is a key-value pair, where both the key and the value

178

# are arbitrary strings that you supply.

179

# For more information, see the documentation on

180

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

181

"a_key": "A String",

182

},

183

"machineType": "A String", # Optional. The type of machine on which to serve the model. Currently only

184

# applies to online prediction service.

185

# <dl>

186

# <dt>mls1-c1-m2</dt>

187

# <dd>

188

# The <b>default</b> machine type, with 1 core and 2 GB RAM. The deprecated

189

# name for this machine type is "mls1-highmem-1".

190

# </dd>

191

# <dt>mls1-c4-m2</dt>

192

# <dd>

193

# In <b>Beta</b>. This machine type has 4 cores and 2 GB RAM. The

194

# deprecated name for this machine type is "mls1-highcpu-4".

195

# </dd>

196

# </dl>

197

"description": "A String", # Optional. The description specified for the version when it was created.

198

"runtimeVersion": "A String", # Optional. The AI Platform runtime version to use for this deployment.

199

# If not set, AI Platform uses the default stable version, 1.0. For more

200

# information, see the

201

# [runtime version list](/ml-engine/docs/runtime-version-list) and

202

# [how to manage runtime versions](/ml-engine/docs/versioning).

203

"manualScaling": { # Options for manually scaling a model. # Manually select the number of nodes to use for serving the

204

# model. You should generally use `auto_scaling` with an appropriate

205

# `min_nodes` instead, but this option is available if you want more

206

# predictable billing. Beware that latency and error rates will increase

207

# if the traffic exceeds that capability of the system to serve it based

208

# on the selected number of nodes.

209

"nodes": 42, # The number of nodes to allocate for this model. These nodes are always up,

210

# starting from the time the model is deployed, so the cost of operating

211

# this model will be proportional to `nodes` * number of hours since

212

# last billing cycle plus the cost for each prediction performed.

213

},

214

"predictionClass": "A String", # Optional. The fully qualified name

215

# (<var>module_name</var>.<var>class_name</var>) of a class that implements

216

# the Predictor interface described in this reference field. The module

217

# containing this class should be included in a package provided to the

218

# [`packageUris` field](#Version.FIELDS.package_uris).

219

#

220

# Specify this field if and only if you are deploying a [custom prediction

221

# routine (beta)](/ml-engine/docs/tensorflow/custom-prediction-routines).

222

# If you specify this field, you must set

223

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

224

#

225

# The following code sample provides the Predictor interface:

226

#

227

# ```py

228

# class Predictor(object):

229

# """Interface for constructing custom predictors."""

230

#

231

# def predict(self, instances, **kwargs):

232

# """Performs custom prediction.

233

#

234

# Instances are the decoded values from the request. They have already

235

# been deserialized from JSON.

236

#

237

# Args:

238

# instances: A list of prediction input instances.

239

# **kwargs: A dictionary of keyword args provided as additional

240

# fields on the predict request body.

241

#

242

# Returns:

243

# A list of outputs containing the prediction results. This list must

244

# be JSON serializable.

245

# """

246

# raise NotImplementedError()

247

#

248

# @classmethod

249

# def from_path(cls, model_dir):

250

# """Creates an instance of Predictor using the given path.

251

#

252

# Loading of the predictor should be done in this method.

253

#

254

# Args:

255

# model_dir: The local directory that contains the exported model

256

# file along with any additional files uploaded when creating the

# version resource.

#

# Returns:

# An instance implementing this Predictor class.

261

# """

262

# raise NotImplementedError()

263

# ```

264

#

265

# Learn more about [the Predictor interface and custom prediction

266

# routines](/ml-engine/docs/tensorflow/custom-prediction-routines).

267

"autoScaling": { # Options for automatically scaling a model. # Automatically scale the number of nodes used to serve the model in

268

# response to increases and decreases in traffic. Care should be

269

# taken to ramp up traffic according to the model's ability to scale

270

# or you will start seeing increases in latency and 429 response codes.

271

"minNodes": 42, # Optional. The minimum number of nodes to allocate for this model. These

272

# nodes are always up, starting from the time the model is deployed.

273

# Therefore, the cost of operating this model will be at least

274

# `rate` * `min_nodes` * number of hours since last billing cycle,

275

# where `rate` is the cost per node-hour as documented in the

276

# [pricing guide](/ml-engine/docs/pricing),

277

# even if no predictions are performed. There is additional cost for each

278

# prediction performed.

279

#

280

# Unlike manual scaling, if the load gets too heavy for the nodes

281

# that are up, the service will automatically add nodes to handle the

282

# increased load as well as scale back as traffic drops, always maintaining

283

# at least `min_nodes`. You will be charged for the time in which additional

284

# nodes are used.

285

#

286

# If not specified, `min_nodes` defaults to 0, in which case, when traffic

287

# to a model stops (and after a cool-down period), nodes will be shut down

288

# and no charges will be incurred until traffic to the model resumes.

289

#

290

# You can set `min_nodes` when creating the model version, and you can also

291

# update `min_nodes` for an existing version:

# <pre>

# update_body.json:

# {

# 'autoScaling': {

# 'minNodes': 5

# }

# }

# </pre>

# HTTP request:

# <pre>

# PATCH

# https://ml.googleapis.com/v1/{name=projects/*/models/*/versions/*}?update_mask=autoScaling.minNodes

304

# -d @./update_body.json

305

# </pre>

306

},

307

"serviceAccount": "A String", # Optional. Specifies the service account for resource access control.

308

"state": "A String", # Output only. The state of a version.

309

"pythonVersion": "A String", # Optional. The version of Python used in prediction. If not set, the default

310

# version is '2.7'. Python '3.5' is available when `runtime_version` is set

311

# to '1.4' and above. Python '2.7' works with all supported runtime versions.

312

"framework": "A String", # Optional. The machine learning framework AI Platform uses to train

313

# this version of the model. Valid values are `TENSORFLOW`, `SCIKIT_LEARN`,

314

# `XGBOOST`. If you do not specify a framework, AI Platform

315

# will analyze files in the deployment_uri to determine a framework. If you

316

# choose `SCIKIT_LEARN` or `XGBOOST`, you must also set the runtime version

317

# of the model to 1.4 or greater.

318

#

319

# Do **not** specify a framework if you're deploying a [custom

320

# prediction routine](/ml-engine/docs/tensorflow/custom-prediction-routines).

321

"packageUris": [ # Optional. Cloud Storage paths (`gs://…`) of packages for [custom

322

# prediction routines](/ml-engine/docs/tensorflow/custom-prediction-routines)

323

# or [scikit-learn pipelines with custom

324

# code](/ml-engine/docs/scikit/exporting-for-prediction#custom-pipeline-code).

325

#

326

# For a custom prediction routine, one of these packages must contain your

327

# Predictor class (see

328

# [`predictionClass`](#Version.FIELDS.prediction_class)). Additionally,

329

# include any dependencies used by your Predictor or scikit-learn pipeline

330

# uses that are not already included in your selected [runtime

331

# version](/ml-engine/docs/tensorflow/runtime-version-list).

332

#

333

# If you specify this field, you must also set

334

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

335

"A String",

336

],

337

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

338

# prevent simultaneous updates of a model from overwriting each other.

339

# It is strongly suggested that systems make use of the `etag` in the

340

# read-modify-write cycle to perform model updates in order to avoid race

341

# conditions: An `etag` is returned in the response to `GetVersion`, and

342

# systems are expected to put that etag in the request to `UpdateVersion` to

343

# ensure that their change will be applied to the model as intended.

344

"lastUseTime": "A String", # Output only. The time the version was last used for prediction.

345

"deploymentUri": "A String", # Required. The Cloud Storage location of the trained model used to

346

# create the version. See the

347

# [guide to model

348

# deployment](/ml-engine/docs/tensorflow/deploying-models) for more

349

# information.

350

#

351

# When passing Version to

352

# [projects.models.versions.create](/ml-engine/reference/rest/v1/projects.models.versions/create)

353

# the model service uses the specified location as the source of the model.

354

# Once deployed, the model version is hosted by the prediction service, so

355

# this location is useful only as a historical record.

356

# The total number of model files can't exceed 1000.

357

"createTime": "A String", # Output only. The time the version was created.

358

"isDefault": True or False, # Output only. If true, this version will be used to handle prediction

359

# requests that do not specify a version.

360

#

361

# You can change the default version by calling

362

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

363

"name": "A String", # Required.The name specified for the version when it was created.

364

#

365

# The version name must be unique within the model it is created in.

366

},

367

"onlinePredictionLogging": True or False, # Optional. If true, online prediction access logs are sent to StackDriver

368

# Logging. These logs are like standard server access logs, containing

369

# information like timestamp and latency for each request. Note that

370

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

371

# your project receives prediction requests at a high queries per second rate

372

# (QPS). Estimate your costs before enabling this option.

373

#

374

# Default is false.

375

"name": "A String", # Required. The name specified for the model when it was created.

376

#

377

# The model name must be unique within the project it is created in.

378

}

379

380

x__xgafv: string, V1 error format.

Allowed values

1 - v1 error format

2 - v2 error format

Returns:

An object of the form:

387

388

{ # Represents a machine learning solution.

389

#

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

390

# A model can have multiple versions, each of which is a deployed, trained

391

# model ready to receive prediction requests. The model itself is just a

392

# container.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

393

"description": "A String", # Optional. The description specified for the model when it was created.

394

"onlinePredictionConsoleLogging": True or False, # Optional. If true, online prediction nodes send `stderr` and `stdout`

395

# streams to Stackdriver Logging. These can be more verbose than the standard

396

# access logs (see `onlinePredictionLogging`) and can incur higher cost.

397

# However, they are helpful for debugging. Note that

398

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

399

# your project receives prediction requests at a high QPS. Estimate your

400

# costs before enabling this option.

401

#

402

# Default is false.

403

"labels": { # Optional. One or more labels that you can add, to organize your models.

404

# Each label is a key-value pair, where both the key and the value are

405

# arbitrary strings that you supply.

406

# For more information, see the documentation on

407

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

408

"a_key": "A String",

409

},

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

410

"regions": [ # Optional. The list of regions where the model is going to be deployed.

411

# Currently only one region per model is supported.

412

# Defaults to 'us-central1' if nothing is set.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

413

# See the <a href="/ml-engine/docs/tensorflow/regions">available regions</a>

414

# for AI Platform services.

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

415

# Note:

416

# * No matter where a model is deployed, it can always be accessed by

417

# users from anywhere, both for online and batch prediction.

418

# * The region for a batch prediction job is set by the region field when

419

# submitting the batch prediction job and does not take its value from

420

# this field.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

421

"A String",

422

],

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

423

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

424

# prevent simultaneous updates of a model from overwriting each other.

425

# It is strongly suggested that systems make use of the `etag` in the

426

# read-modify-write cycle to perform model updates in order to avoid race

427

# conditions: An `etag` is returned in the response to `GetModel`, and

428

# systems are expected to put that etag in the request to `UpdateModel` to

429

# ensure that their change will be applied to the model as intended.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

430

"defaultVersion": { # Represents a version of the model. # Output only. The default version of the model. This version will be used to

431

# handle prediction requests that do not specify a version.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

432

#

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

433

# You can change the default version by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

434

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

435

#

436

# Each version is a trained model deployed in the cloud, ready to handle

437

# prediction requests. A model can have multiple versions. You can get

438

# information about all of the versions of a given model by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

439

# [projects.models.versions.list](/ml-engine/reference/rest/v1/projects.models.versions/list).

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

440

"errorMessage": "A String", # Output only. The details of a failure or a cancellation.

441

"labels": { # Optional. One or more labels that you can add, to organize your model

442

# versions. Each label is a key-value pair, where both the key and the value

443

# are arbitrary strings that you supply.

444

# For more information, see the documentation on

445

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

446

"a_key": "A String",

447

},

448

"machineType": "A String", # Optional. The type of machine on which to serve the model. Currently only

449

# applies to online prediction service.

450

# <dl>

451

# <dt>mls1-c1-m2</dt>

452

# <dd>

453

# The <b>default</b> machine type, with 1 core and 2 GB RAM. The deprecated

454

# name for this machine type is "mls1-highmem-1".

455

# </dd>

456

# <dt>mls1-c4-m2</dt>

457

# <dd>

458

# In <b>Beta</b>. This machine type has 4 cores and 2 GB RAM. The

459

# deprecated name for this machine type is "mls1-highcpu-4".

460

# </dd>

461

# </dl>

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

462

"description": "A String", # Optional. The description specified for the version when it was created.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

463

"runtimeVersion": "A String", # Optional. The AI Platform runtime version to use for this deployment.

464

# If not set, AI Platform uses the default stable version, 1.0. For more

465

# information, see the

466

# [runtime version list](/ml-engine/docs/runtime-version-list) and

467

# [how to manage runtime versions](/ml-engine/docs/versioning).

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

468

"manualScaling": { # Options for manually scaling a model. # Manually select the number of nodes to use for serving the

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

469

# model. You should generally use `auto_scaling` with an appropriate

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

470

# `min_nodes` instead, but this option is available if you want more

471

# predictable billing. Beware that latency and error rates will increase

472

# if the traffic exceeds that capability of the system to serve it based

473

# on the selected number of nodes.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

474

"nodes": 42, # The number of nodes to allocate for this model. These nodes are always up,

475

# starting from the time the model is deployed, so the cost of operating

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

476

# this model will be proportional to `nodes` * number of hours since

477

# last billing cycle plus the cost for each prediction performed.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

478

},

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

479

"predictionClass": "A String", # Optional. The fully qualified name

480

# (<var>module_name</var>.<var>class_name</var>) of a class that implements

481

# the Predictor interface described in this reference field. The module

482

# containing this class should be included in a package provided to the

483

# [`packageUris` field](#Version.FIELDS.package_uris).

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

484

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

485

# Specify this field if and only if you are deploying a [custom prediction

486

# routine (beta)](/ml-engine/docs/tensorflow/custom-prediction-routines).

487

# If you specify this field, you must set

488

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

489

#

490

# The following code sample provides the Predictor interface:

491

#

492

# ```py

493

# class Predictor(object):

494

# """Interface for constructing custom predictors."""

495

#

496

# def predict(self, instances, **kwargs):

497

# """Performs custom prediction.

498

#

499

# Instances are the decoded values from the request. They have already

500

# been deserialized from JSON.

501

#

502

# Args:

503

# instances: A list of prediction input instances.

504

# **kwargs: A dictionary of keyword args provided as additional

505

# fields on the predict request body.

506

#

507

# Returns:

508

# A list of outputs containing the prediction results. This list must

509

# be JSON serializable.

510

# """

511

# raise NotImplementedError()

512

#

513

# @classmethod

514

# def from_path(cls, model_dir):

515

# """Creates an instance of Predictor using the given path.

516

#

517

# Loading of the predictor should be done in this method.

518

#

519

# Args:

520

# model_dir: The local directory that contains the exported model

521

# file along with any additional files uploaded when creating the

# version resource.

#

# Returns:

# An instance implementing this Predictor class.

526

# """

527

# raise NotImplementedError()

528

# ```

529

#

530

# Learn more about [the Predictor interface and custom prediction

531

# routines](/ml-engine/docs/tensorflow/custom-prediction-routines).

532

"autoScaling": { # Options for automatically scaling a model. # Automatically scale the number of nodes used to serve the model in

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

533

# response to increases and decreases in traffic. Care should be

534

# taken to ramp up traffic according to the model's ability to scale

535

# or you will start seeing increases in latency and 429 response codes.

536

"minNodes": 42, # Optional. The minimum number of nodes to allocate for this model. These

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

537

# nodes are always up, starting from the time the model is deployed.

538

# Therefore, the cost of operating this model will be at least

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

539

# `rate` * `min_nodes` * number of hours since last billing cycle,

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

540

# where `rate` is the cost per node-hour as documented in the

541

# [pricing guide](/ml-engine/docs/pricing),

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

542

# even if no predictions are performed. There is additional cost for each

543

# prediction performed.

544

#

545

# Unlike manual scaling, if the load gets too heavy for the nodes

546

# that are up, the service will automatically add nodes to handle the

547

# increased load as well as scale back as traffic drops, always maintaining

548

# at least `min_nodes`. You will be charged for the time in which additional

549

# nodes are used.

550

#

551

# If not specified, `min_nodes` defaults to 0, in which case, when traffic

552

# to a model stops (and after a cool-down period), nodes will be shut down

553

# and no charges will be incurred until traffic to the model resumes.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

554

#

555

# You can set `min_nodes` when creating the model version, and you can also

556

# update `min_nodes` for an existing version:

# <pre>

# update_body.json:

# {

# 'autoScaling': {

# 'minNodes': 5

# }

# }

# </pre>

# HTTP request:

# <pre>

# PATCH

# https://ml.googleapis.com/v1/{name=projects/*/models/*/versions/*}?update_mask=autoScaling.minNodes

569

# -d @./update_body.json

570

# </pre>

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

571

},

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

572

"serviceAccount": "A String", # Optional. Specifies the service account for resource access control.

573

"state": "A String", # Output only. The state of a version.

574

"pythonVersion": "A String", # Optional. The version of Python used in prediction. If not set, the default

575

# version is '2.7'. Python '3.5' is available when `runtime_version` is set

576

# to '1.4' and above. Python '2.7' works with all supported runtime versions.

577

"framework": "A String", # Optional. The machine learning framework AI Platform uses to train

578

# this version of the model. Valid values are `TENSORFLOW`, `SCIKIT_LEARN`,

579

# `XGBOOST`. If you do not specify a framework, AI Platform

580

# will analyze files in the deployment_uri to determine a framework. If you

581

# choose `SCIKIT_LEARN` or `XGBOOST`, you must also set the runtime version

582

# of the model to 1.4 or greater.

583

#

584

# Do **not** specify a framework if you're deploying a [custom

585

# prediction routine](/ml-engine/docs/tensorflow/custom-prediction-routines).

586

"packageUris": [ # Optional. Cloud Storage paths (`gs://…`) of packages for [custom

587

# prediction routines](/ml-engine/docs/tensorflow/custom-prediction-routines)

588

# or [scikit-learn pipelines with custom

589

# code](/ml-engine/docs/scikit/exporting-for-prediction#custom-pipeline-code).

590

#

591

# For a custom prediction routine, one of these packages must contain your

592

# Predictor class (see

593

# [`predictionClass`](#Version.FIELDS.prediction_class)). Additionally,

594

# include any dependencies used by your Predictor or scikit-learn pipeline

595

# uses that are not already included in your selected [runtime

596

# version](/ml-engine/docs/tensorflow/runtime-version-list).

597

#

598

# If you specify this field, you must also set

599

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

600

"A String",

601

],

602

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

603

# prevent simultaneous updates of a model from overwriting each other.

604

# It is strongly suggested that systems make use of the `etag` in the

605

# read-modify-write cycle to perform model updates in order to avoid race

606

# conditions: An `etag` is returned in the response to `GetVersion`, and

607

# systems are expected to put that etag in the request to `UpdateVersion` to

608

# ensure that their change will be applied to the model as intended.

609

"lastUseTime": "A String", # Output only. The time the version was last used for prediction.

610

"deploymentUri": "A String", # Required. The Cloud Storage location of the trained model used to

611

# create the version. See the

612

# [guide to model

613

# deployment](/ml-engine/docs/tensorflow/deploying-models) for more

614

# information.

615

#

616

# When passing Version to

617

# [projects.models.versions.create](/ml-engine/reference/rest/v1/projects.models.versions/create)

618

# the model service uses the specified location as the source of the model.

619

# Once deployed, the model version is hosted by the prediction service, so

620

# this location is useful only as a historical record.

621

# The total number of model files can't exceed 1000.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

622

"createTime": "A String", # Output only. The time the version was created.

623

"isDefault": True or False, # Output only. If true, this version will be used to handle prediction

624

# requests that do not specify a version.

625

#

626

# You can change the default version by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

627

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

628

"name": "A String", # Required.The name specified for the version when it was created.

629

#

630

# The version name must be unique within the model it is created in.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

631

},

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

632

"onlinePredictionLogging": True or False, # Optional. If true, online prediction access logs are sent to StackDriver

633

# Logging. These logs are like standard server access logs, containing

634

# information like timestamp and latency for each request. Note that

635

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

636

# your project receives prediction requests at a high queries per second rate

637

# (QPS). Estimate your costs before enabling this option.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

638

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

639

# Default is false.

640

"name": "A String", # Required. The name specified for the model when it was created.

641

#

642

# The model name must be unique within the project it is created in.

643

}</pre>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

</div>

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

647

<code class="details" id="delete">delete(name, x__xgafv=None)</code>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

648

<pre>Deletes a model.

649

650

You can only delete a model if there are no versions in it. You can delete

651

versions by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

652

[projects.models.versions.delete](/ml-engine/reference/rest/v1/projects.models.versions/delete).

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

653

654

Args:

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

655

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

656

x__xgafv: string, V1 error format.

Allowed values

1 - v1 error format

2 - v2 error format

Returns:

An object of the form:

663

664

{ # This resource represents a long-running operation that is the result of a

665

# network API call.

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

666

"metadata": { # Service-specific metadata associated with the operation. It typically

667

# contains progress information and common metadata such as create time.

668

# Some services might not provide such metadata. Any method that returns a

669

# long-running operation should document the metadata type, if any.

670

"a_key": "", # Properties of the object. Contains field @type with type URL.

671

},

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

672

"error": { # The `Status` type defines a logical error model that is suitable for # The error result of the operation in case of failure or cancellation.

673

# different programming environments, including REST APIs and RPC APIs. It is

674

# used by [gRPC](https://github.com/grpc). Each `Status` message contains

675

# three pieces of data: error code, error message, and error details.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

676

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

677

# You can find out more about this error model and how to work with it in the

678

# [API Design Guide](https://cloud.google.com/apis/design/errors).

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

679

"message": "A String", # A developer-facing error message, which should be in English. Any

680

# user-facing error message should be localized and sent in the

681

# google.rpc.Status.details field, or localized by the client.

682

"code": 42, # The status code, which should be an enum value of google.rpc.Code.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

683

"details": [ # A list of messages that carry the error details. There is a common set of

684

# message types for APIs to use.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

685

{

686

"a_key": "", # Properties of the object. Contains field @type with type URL.

687

},

688

],

689

},

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

690

"done": True or False, # If the value is `false`, it means the operation is still in progress.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

691

# If `true`, the operation is completed, and either `error` or `response` is

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

692

# available.

693

"response": { # The normal response of the operation in case of success. If the original

694

# method returns no data on success, such as `Delete`, the response is

695

# `google.protobuf.Empty`. If the original method is standard

696

# `Get`/`Create`/`Update`, the response should be the resource. For other

697

# methods, the response should have the type `XxxResponse`, where `Xxx`

698

# is the original method name. For example, if the original method name

699

# is `TakeSnapshot()`, the inferred response type is

700

# `TakeSnapshotResponse`.

701

"a_key": "", # Properties of the object. Contains field @type with type URL.

702

},

703

"name": "A String", # The server-assigned name, which is only unique within the same service that

704

# originally returns it. If you use the default HTTP mapping, the

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

705

# `name` should be a resource name ending with `operations/{unique_id}`.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

}</pre>

</div>

Thomas Coffee

2017-03-27 10:39:26 -0700

[diff] [blame]

710

<code class="details" id="get">get(name, x__xgafv=None)</code>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

711

<pre>Gets information about a model, including its name, the description (if

712

set), and the default version (if at least one version of the model has

713

been deployed).

714

715

Args:

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

716

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

717

x__xgafv: string, V1 error format.

Allowed values

1 - v1 error format

2 - v2 error format

Returns:

An object of the form:

724

725

{ # Represents a machine learning solution.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

726

#

727

# A model can have multiple versions, each of which is a deployed, trained

728

# model ready to receive prediction requests. The model itself is just a

729

# container.

730

"description": "A String", # Optional. The description specified for the model when it was created.

731

"onlinePredictionConsoleLogging": True or False, # Optional. If true, online prediction nodes send `stderr` and `stdout`

732

# streams to Stackdriver Logging. These can be more verbose than the standard

733

# access logs (see `onlinePredictionLogging`) and can incur higher cost.

734

# However, they are helpful for debugging. Note that

735

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

736

# your project receives prediction requests at a high QPS. Estimate your

737

# costs before enabling this option.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

738

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

739

# Default is false.

740

"labels": { # Optional. One or more labels that you can add, to organize your models.

741

# Each label is a key-value pair, where both the key and the value are

742

# arbitrary strings that you supply.

743

# For more information, see the documentation on

744

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

745

"a_key": "A String",

746

},

747

"regions": [ # Optional. The list of regions where the model is going to be deployed.

748

# Currently only one region per model is supported.

749

# Defaults to 'us-central1' if nothing is set.

750

# See the <a href="/ml-engine/docs/tensorflow/regions">available regions</a>

751

# for AI Platform services.

752

# Note:

753

# * No matter where a model is deployed, it can always be accessed by

754

# users from anywhere, both for online and batch prediction.

755

# * The region for a batch prediction job is set by the region field when

756

# submitting the batch prediction job and does not take its value from

# this field.

"A String",

],

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

761

# prevent simultaneous updates of a model from overwriting each other.

762

# It is strongly suggested that systems make use of the `etag` in the

763

# read-modify-write cycle to perform model updates in order to avoid race

764

# conditions: An `etag` is returned in the response to `GetModel`, and

765

# systems are expected to put that etag in the request to `UpdateModel` to

766

# ensure that their change will be applied to the model as intended.

767

"defaultVersion": { # Represents a version of the model. # Output only. The default version of the model. This version will be used to

768

# handle prediction requests that do not specify a version.

769

#

770

# You can change the default version by calling

771

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

772

#

773

# Each version is a trained model deployed in the cloud, ready to handle

774

# prediction requests. A model can have multiple versions. You can get

775

# information about all of the versions of a given model by calling

776

# [projects.models.versions.list](/ml-engine/reference/rest/v1/projects.models.versions/list).

777

"errorMessage": "A String", # Output only. The details of a failure or a cancellation.

778

"labels": { # Optional. One or more labels that you can add, to organize your model

779

# versions. Each label is a key-value pair, where both the key and the value

780

# are arbitrary strings that you supply.

781

# For more information, see the documentation on

782

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

783

"a_key": "A String",

784

},

785

"machineType": "A String", # Optional. The type of machine on which to serve the model. Currently only

786

# applies to online prediction service.

787

# <dl>

788

# <dt>mls1-c1-m2</dt>

789

# <dd>

790

# The <b>default</b> machine type, with 1 core and 2 GB RAM. The deprecated

791

# name for this machine type is "mls1-highmem-1".

792

# </dd>

793

# <dt>mls1-c4-m2</dt>

794

# <dd>

795

# In <b>Beta</b>. This machine type has 4 cores and 2 GB RAM. The

796

# deprecated name for this machine type is "mls1-highcpu-4".

797

# </dd>

798

# </dl>

799

"description": "A String", # Optional. The description specified for the version when it was created.

800

"runtimeVersion": "A String", # Optional. The AI Platform runtime version to use for this deployment.

801

# If not set, AI Platform uses the default stable version, 1.0. For more

802

# information, see the

803

# [runtime version list](/ml-engine/docs/runtime-version-list) and

804

# [how to manage runtime versions](/ml-engine/docs/versioning).

805

"manualScaling": { # Options for manually scaling a model. # Manually select the number of nodes to use for serving the

806

# model. You should generally use `auto_scaling` with an appropriate

807

# `min_nodes` instead, but this option is available if you want more

808

# predictable billing. Beware that latency and error rates will increase

809

# if the traffic exceeds that capability of the system to serve it based

810

# on the selected number of nodes.

811

"nodes": 42, # The number of nodes to allocate for this model. These nodes are always up,

812

# starting from the time the model is deployed, so the cost of operating

813

# this model will be proportional to `nodes` * number of hours since

814

# last billing cycle plus the cost for each prediction performed.

815

},

816

"predictionClass": "A String", # Optional. The fully qualified name

817

# (<var>module_name</var>.<var>class_name</var>) of a class that implements

818

# the Predictor interface described in this reference field. The module

819

# containing this class should be included in a package provided to the

820

# [`packageUris` field](#Version.FIELDS.package_uris).

821

#

822

# Specify this field if and only if you are deploying a [custom prediction

823

# routine (beta)](/ml-engine/docs/tensorflow/custom-prediction-routines).

824

# If you specify this field, you must set

825

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

826

#

827

# The following code sample provides the Predictor interface:

828

#

829

# ```py

830

# class Predictor(object):

831

# """Interface for constructing custom predictors."""

832

#

833

# def predict(self, instances, **kwargs):

834

# """Performs custom prediction.

835

#

836

# Instances are the decoded values from the request. They have already

837

# been deserialized from JSON.

838

#

839

# Args:

840

# instances: A list of prediction input instances.

841

# **kwargs: A dictionary of keyword args provided as additional

842

# fields on the predict request body.

843

#

844

# Returns:

845

# A list of outputs containing the prediction results. This list must

846

# be JSON serializable.

847

# """

848

# raise NotImplementedError()

849

#

850

# @classmethod

851

# def from_path(cls, model_dir):

852

# """Creates an instance of Predictor using the given path.

853

#

854

# Loading of the predictor should be done in this method.

855

#

856

# Args:

857

# model_dir: The local directory that contains the exported model

858

# file along with any additional files uploaded when creating the

# version resource.

#

# Returns:

# An instance implementing this Predictor class.

863

# """

864

# raise NotImplementedError()

865

# ```

866

#

867

# Learn more about [the Predictor interface and custom prediction

868

# routines](/ml-engine/docs/tensorflow/custom-prediction-routines).

869

"autoScaling": { # Options for automatically scaling a model. # Automatically scale the number of nodes used to serve the model in

870

# response to increases and decreases in traffic. Care should be

871

# taken to ramp up traffic according to the model's ability to scale

872

# or you will start seeing increases in latency and 429 response codes.

873

"minNodes": 42, # Optional. The minimum number of nodes to allocate for this model. These

874

# nodes are always up, starting from the time the model is deployed.

875

# Therefore, the cost of operating this model will be at least

876

# `rate` * `min_nodes` * number of hours since last billing cycle,

877

# where `rate` is the cost per node-hour as documented in the

878

# [pricing guide](/ml-engine/docs/pricing),

879

# even if no predictions are performed. There is additional cost for each

880

# prediction performed.

881

#

882

# Unlike manual scaling, if the load gets too heavy for the nodes

883

# that are up, the service will automatically add nodes to handle the

884

# increased load as well as scale back as traffic drops, always maintaining

885

# at least `min_nodes`. You will be charged for the time in which additional

886

# nodes are used.

887

#

888

# If not specified, `min_nodes` defaults to 0, in which case, when traffic

889

# to a model stops (and after a cool-down period), nodes will be shut down

890

# and no charges will be incurred until traffic to the model resumes.

891

#

892

# You can set `min_nodes` when creating the model version, and you can also

893

# update `min_nodes` for an existing version:

# <pre>

# update_body.json:

# {

# 'autoScaling': {

# 'minNodes': 5

# }

# }

# </pre>

# HTTP request:

# <pre>

# PATCH

# https://ml.googleapis.com/v1/{name=projects/*/models/*/versions/*}?update_mask=autoScaling.minNodes

906

# -d @./update_body.json

907

# </pre>

908

},

909

"serviceAccount": "A String", # Optional. Specifies the service account for resource access control.

910

"state": "A String", # Output only. The state of a version.

911

"pythonVersion": "A String", # Optional. The version of Python used in prediction. If not set, the default

912

# version is '2.7'. Python '3.5' is available when `runtime_version` is set

913

# to '1.4' and above. Python '2.7' works with all supported runtime versions.

914

"framework": "A String", # Optional. The machine learning framework AI Platform uses to train

915

# this version of the model. Valid values are `TENSORFLOW`, `SCIKIT_LEARN`,

916

# `XGBOOST`. If you do not specify a framework, AI Platform

917

# will analyze files in the deployment_uri to determine a framework. If you

918

# choose `SCIKIT_LEARN` or `XGBOOST`, you must also set the runtime version

919

# of the model to 1.4 or greater.

920

#

921

# Do **not** specify a framework if you're deploying a [custom

922

# prediction routine](/ml-engine/docs/tensorflow/custom-prediction-routines).

923

"packageUris": [ # Optional. Cloud Storage paths (`gs://…`) of packages for [custom

924

# prediction routines](/ml-engine/docs/tensorflow/custom-prediction-routines)

925

# or [scikit-learn pipelines with custom

926

# code](/ml-engine/docs/scikit/exporting-for-prediction#custom-pipeline-code).

927

#

928

# For a custom prediction routine, one of these packages must contain your

929

# Predictor class (see

930

# [`predictionClass`](#Version.FIELDS.prediction_class)). Additionally,

931

# include any dependencies used by your Predictor or scikit-learn pipeline

932

# uses that are not already included in your selected [runtime

933

# version](/ml-engine/docs/tensorflow/runtime-version-list).

934

#

935

# If you specify this field, you must also set

936

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

937

"A String",

938

],

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

939

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

940

# prevent simultaneous updates of a model from overwriting each other.

941

# It is strongly suggested that systems make use of the `etag` in the

942

# read-modify-write cycle to perform model updates in order to avoid race

943

# conditions: An `etag` is returned in the response to `GetVersion`, and

944

# systems are expected to put that etag in the request to `UpdateVersion` to

945

# ensure that their change will be applied to the model as intended.

946

"lastUseTime": "A String", # Output only. The time the version was last used for prediction.

947

"deploymentUri": "A String", # Required. The Cloud Storage location of the trained model used to

948

# create the version. See the

949

# [guide to model

950

# deployment](/ml-engine/docs/tensorflow/deploying-models) for more

951

# information.

952

#

953

# When passing Version to

954

# [projects.models.versions.create](/ml-engine/reference/rest/v1/projects.models.versions/create)

955

# the model service uses the specified location as the source of the model.

956

# Once deployed, the model version is hosted by the prediction service, so

957

# this location is useful only as a historical record.

958

# The total number of model files can't exceed 1000.

959

"createTime": "A String", # Output only. The time the version was created.

960

"isDefault": True or False, # Output only. If true, this version will be used to handle prediction

961

# requests that do not specify a version.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

962

#

963

# You can change the default version by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

964

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

965

"name": "A String", # Required.The name specified for the version when it was created.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

966

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

967

# The version name must be unique within the model it is created in.

968

},

969

"onlinePredictionLogging": True or False, # Optional. If true, online prediction access logs are sent to StackDriver

970

# Logging. These logs are like standard server access logs, containing

971

# information like timestamp and latency for each request. Note that

972

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

973

# your project receives prediction requests at a high queries per second rate

974

# (QPS). Estimate your costs before enabling this option.

975

#

976

# Default is false.

977

"name": "A String", # Required. The name specified for the model when it was created.

978

#

979

# The model name must be unique within the project it is created in.

980

}</pre>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

</div>

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

984

<code class="details" id="getIamPolicy">getIamPolicy(resource, x__xgafv=None)</code>

985

<pre>Gets the access control policy for a resource.

986

Returns an empty policy if the resource exists and does not have a policy

set.

Args:

resource: string, REQUIRED: The resource for which the policy is being requested.

991

See the operation documentation for the appropriate value for this field. (required)

992

x__xgafv: string, V1 error format.

Allowed values

1 - v1 error format

2 - v2 error format

Returns:

An object of the form:

999

1000

{ # Defines an Identity and Access Management (IAM) policy. It is used to

1001

# specify access control policies for Cloud Platform resources.

1002

#

1003

#

1004

# A `Policy` consists of a list of `bindings`. A `binding` binds a list of

1005

# `members` to a `role`, where the members can be user accounts, Google groups,

1006

# Google domains, and service accounts. A `role` is a named list of permissions

# defined by IAM.

#

# **JSON Example**

#

# {

# "bindings": [

# {

# "role": "roles/owner",

1015

# "members": [

1016

# "user:mike@example.com",

1017

# "group:admins@example.com",

1018

# "domain:google.com",

1019

# "serviceAccount:my-other-app@appspot.gserviceaccount.com"

# ]

# },

# {

# "role": "roles/viewer",

1024

# "members": ["user:sean@example.com"]

# }

# ]

# }

#

# **YAML Example**

#

# bindings:

# - members:

# - user:mike@example.com

1034

# - group:admins@example.com

1035

# - domain:google.com

1036

# - serviceAccount:my-other-app@appspot.gserviceaccount.com

1037

# role: roles/owner

1038

# - members:

1039

# - user:sean@example.com

# role: roles/viewer

#

#

# For a description of IAM and its features, see the

1044

# [IAM developer's guide](https://cloud.google.com/iam/docs).

1045

"bindings": [ # Associates a list of `members` to a `role`.

1046

# `bindings` with no members will result in an error.

1047

{ # Associates `members` with a `role`.

1048

"role": "A String", # Role that is assigned to `members`.

1049

# For example, `roles/viewer`, `roles/editor`, or `roles/owner`.

1050

"members": [ # Specifies the identities requesting access for a Cloud Platform resource.

1051

# `members` can have the following values:

1052

#

1053

# * `allUsers`: A special identifier that represents anyone who is

1054

# on the internet; with or without a Google account.

1055

#

1056

# * `allAuthenticatedUsers`: A special identifier that represents anyone

1057

# who is authenticated with a Google account or a service account.

1058

#

1059

# * `user:{emailid}`: An email address that represents a specific Google

1060

# account. For example, `alice@gmail.com` .

1061

#

1062

#

1063

# * `serviceAccount:{emailid}`: An email address that represents a service

1064

# account. For example, `my-other-app@appspot.gserviceaccount.com`.

1065

#

1066

# * `group:{emailid}`: An email address that represents a Google group.

1067

# For example, `admins@example.com`.

1068

#

1069

#

1070

# * `domain:{domain}`: The G Suite domain (primary) that represents all the

1071

# users of that domain. For example, `google.com` or `example.com`.

#

"A String",

],

"condition": { # Represents an expression text. Example: # The condition that is associated with this binding.

1076

# NOTE: An unsatisfied condition will not allow user access via current

1077

# binding. Different bindings, including their conditions, are examined

1078

# independently.

1079

#

1080

# title: "User account presence"

1081

# description: "Determines whether the request has a user account"

1082

# expression: "size(request.user) > 0"

1083

"description": "A String", # An optional description of the expression. This is a longer text which

1084

# describes the expression, e.g. when hovered over it in a UI.

1085

"expression": "A String", # Textual representation of an expression in

1086

# Common Expression Language syntax.

1087

#

1088

# The application context of the containing message determines which

1089

# well-known feature set of CEL is supported.

1090

"location": "A String", # An optional string indicating the location of the expression for error

1091

# reporting, e.g. a file name and a position in the file.

1092

"title": "A String", # An optional title for the expression, i.e. a short string describing

1093

# its purpose. This can be used e.g. in UIs which allow to enter the

# expression.

},

},

],

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

1099

# prevent simultaneous updates of a policy from overwriting each other.

1100

# It is strongly suggested that systems make use of the `etag` in the

1101

# read-modify-write cycle to perform policy updates in order to avoid race

1102

# conditions: An `etag` is returned in the response to `getIamPolicy`, and

1103

# systems are expected to put that etag in the request to `setIamPolicy` to

1104

# ensure that their change will be applied to the same version of the policy.

1105

#

1106

# If no `etag` is provided in the call to `setIamPolicy`, then the existing

1107

# policy is overwritten blindly.

1108

"version": 42, # Deprecated.

1109

"auditConfigs": [ # Specifies cloud audit logging configuration for this policy.

1110

{ # Specifies the audit configuration for a service.

1111

# The configuration determines which permission types are logged, and what

1112

# identities, if any, are exempted from logging.

1113

# An AuditConfig must have one or more AuditLogConfigs.

1114

#

1115

# If there are AuditConfigs for both `allServices` and a specific service,

1116

# the union of the two AuditConfigs is used for that service: the log_types

1117

# specified in each AuditConfig are enabled, and the exempted_members in each

1118

# AuditLogConfig are exempted.

1119

#

1120

# Example Policy with multiple AuditConfigs:

#

# {

# "audit_configs": [

# {

# "service": "allServices"

1126

# "audit_log_configs": [

1127

# {

1128

# "log_type": "DATA_READ",

1129

# "exempted_members": [

1130

# "user:foo@gmail.com"

# ]

# },

# {

# "log_type": "DATA_WRITE",

1135

# },

1136

# {

1137

# "log_type": "ADMIN_READ",

# }

# ]

# },

# {

# "service": "fooservice.googleapis.com"

1143

# "audit_log_configs": [

1144

# {

1145

# "log_type": "DATA_READ",

1146

# },

1147

# {

1148

# "log_type": "DATA_WRITE",

1149

# "exempted_members": [

1150

# "user:bar@gmail.com"

# ]

# }

# ]

# }

# ]

# }

#

# For fooservice, this policy enables DATA_READ, DATA_WRITE and ADMIN_READ

1159

# logging. It also exempts foo@gmail.com from DATA_READ logging, and

1160

# bar@gmail.com from DATA_WRITE logging.

1161

"auditLogConfigs": [ # The configuration for logging of each type of permission.

1162

{ # Provides the configuration for logging a type of permissions.

# Example:

#

# {

# "audit_log_configs": [

1167

# {

1168

# "log_type": "DATA_READ",

1169

# "exempted_members": [

1170

# "user:foo@gmail.com"

# ]

# },

# {

# "log_type": "DATA_WRITE",

# }

# ]

# }

#

# This enables 'DATA_READ' and 'DATA_WRITE' logging, while exempting

1180

# foo@gmail.com from DATA_READ logging.

1181

"exemptedMembers": [ # Specifies the identities that do not cause logging for this type of

1182

# permission.

1183

# Follows the same format of Binding.members.

1184

"A String",

1185

],

1186

"logType": "A String", # The log type that this config enables.

1187

},

1188

],

1189

"service": "A String", # Specifies a service that will be enabled for audit logging.

1190

# For example, `storage.googleapis.com`, `cloudsql.googleapis.com`.

1191

# `allServices` is a special value that covers all services.

},

],

}</pre>

</div>

<code class="details" id="list">list(parent, pageToken=None, x__xgafv=None, pageSize=None, filter=None)</code>

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1199

<pre>Lists the models in a project.

1200

1201

Each project can contain multiple models, and each model can have multiple

1202

versions.

1203

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1204

If there are no models that match the request parameters, the list request

1205

returns an empty response body: {}.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1206

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1207

Args:

1208

parent: string, Required. The name of the project whose models are to be listed. (required)

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1209

pageToken: string, Optional. A page token to request the next page of results.

1210

1211

You get the token from the `next_page_token` field of the response from

1212

the previous call.

1213

x__xgafv: string, V1 error format.

1214

Allowed values

1215

1 - v1 error format

1216

2 - v2 error format

Sai Cheemalapati

2017-06-06 18:46:08 -0400

[diff] [blame]

1217

pageSize: integer, Optional. The number of models to retrieve per "page" of results. If there

1218

are more remaining results than this number, the response message will

1219

contain a valid value in the `next_page_token` field.

1220

1221

The default value is 20, and the maximum page size is 100.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1222

filter: string, Optional. Specifies the subset of models to retrieve.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1223

1224

Returns:

1225

An object of the form:

1226

1227

{ # Response message for the ListModels method.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1228

"nextPageToken": "A String", # Optional. Pass this token as the `page_token` field of the request for a

1229

# subsequent call.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1230

"models": [ # The list of models.

1231

{ # Represents a machine learning solution.

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1232

#

1233

# A model can have multiple versions, each of which is a deployed, trained

1234

# model ready to receive prediction requests. The model itself is just a

1235

# container.

1236

"description": "A String", # Optional. The description specified for the model when it was created.

1237

"onlinePredictionConsoleLogging": True or False, # Optional. If true, online prediction nodes send `stderr` and `stdout`

1238

# streams to Stackdriver Logging. These can be more verbose than the standard

1239

# access logs (see `onlinePredictionLogging`) and can incur higher cost.

1240

# However, they are helpful for debugging. Note that

1241

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

1242

# your project receives prediction requests at a high QPS. Estimate your

1243

# costs before enabling this option.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1244

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1245

# Default is false.

1246

"labels": { # Optional. One or more labels that you can add, to organize your models.

1247

# Each label is a key-value pair, where both the key and the value are

1248

# arbitrary strings that you supply.

1249

# For more information, see the documentation on

1250

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

1251

"a_key": "A String",

1252

},

1253

"regions": [ # Optional. The list of regions where the model is going to be deployed.

1254

# Currently only one region per model is supported.

1255

# Defaults to 'us-central1' if nothing is set.

1256

# See the <a href="/ml-engine/docs/tensorflow/regions">available regions</a>

1257

# for AI Platform services.

1258

# Note:

1259

# * No matter where a model is deployed, it can always be accessed by

1260

# users from anywhere, both for online and batch prediction.

1261

# * The region for a batch prediction job is set by the region field when

1262

# submitting the batch prediction job and does not take its value from

# this field.

"A String",

],

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

1267

# prevent simultaneous updates of a model from overwriting each other.

1268

# It is strongly suggested that systems make use of the `etag` in the

1269

# read-modify-write cycle to perform model updates in order to avoid race

1270

# conditions: An `etag` is returned in the response to `GetModel`, and

1271

# systems are expected to put that etag in the request to `UpdateModel` to

1272

# ensure that their change will be applied to the model as intended.

1273

"defaultVersion": { # Represents a version of the model. # Output only. The default version of the model. This version will be used to

1274

# handle prediction requests that do not specify a version.

1275

#

1276

# You can change the default version by calling

1277

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

1278

#

1279

# Each version is a trained model deployed in the cloud, ready to handle

1280

# prediction requests. A model can have multiple versions. You can get

1281

# information about all of the versions of a given model by calling

1282

# [projects.models.versions.list](/ml-engine/reference/rest/v1/projects.models.versions/list).

1283

"errorMessage": "A String", # Output only. The details of a failure or a cancellation.

1284

"labels": { # Optional. One or more labels that you can add, to organize your model

1285

# versions. Each label is a key-value pair, where both the key and the value

1286

# are arbitrary strings that you supply.

1287

# For more information, see the documentation on

1288

# <a href="/ml-engine/docs/tensorflow/resource-labels">using labels</a>.

1289

"a_key": "A String",

1290

},

1291

"machineType": "A String", # Optional. The type of machine on which to serve the model. Currently only

1292

# applies to online prediction service.

1293

# <dl>

1294

# <dt>mls1-c1-m2</dt>

1295

# <dd>

1296

# The <b>default</b> machine type, with 1 core and 2 GB RAM. The deprecated

1297

# name for this machine type is "mls1-highmem-1".

1298

# </dd>

1299

# <dt>mls1-c4-m2</dt>

1300

# <dd>

1301

# In <b>Beta</b>. This machine type has 4 cores and 2 GB RAM. The

1302

# deprecated name for this machine type is "mls1-highcpu-4".

1303

# </dd>

1304

# </dl>

1305

"description": "A String", # Optional. The description specified for the version when it was created.

1306

"runtimeVersion": "A String", # Optional. The AI Platform runtime version to use for this deployment.

1307

# If not set, AI Platform uses the default stable version, 1.0. For more

1308

# information, see the

1309

# [runtime version list](/ml-engine/docs/runtime-version-list) and

1310

# [how to manage runtime versions](/ml-engine/docs/versioning).

1311

"manualScaling": { # Options for manually scaling a model. # Manually select the number of nodes to use for serving the

1312

# model. You should generally use `auto_scaling` with an appropriate

1313

# `min_nodes` instead, but this option is available if you want more

1314

# predictable billing. Beware that latency and error rates will increase

1315

# if the traffic exceeds that capability of the system to serve it based

1316

# on the selected number of nodes.

1317

"nodes": 42, # The number of nodes to allocate for this model. These nodes are always up,

1318

# starting from the time the model is deployed, so the cost of operating

1319

# this model will be proportional to `nodes` * number of hours since

1320

# last billing cycle plus the cost for each prediction performed.

1321

},

1322

"predictionClass": "A String", # Optional. The fully qualified name

1323

# (<var>module_name</var>.<var>class_name</var>) of a class that implements

1324

# the Predictor interface described in this reference field. The module

1325

# containing this class should be included in a package provided to the

1326

# [`packageUris` field](#Version.FIELDS.package_uris).

1327

#

1328

# Specify this field if and only if you are deploying a [custom prediction

1329

# routine (beta)](/ml-engine/docs/tensorflow/custom-prediction-routines).

1330

# If you specify this field, you must set

1331

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

1332

#

1333

# The following code sample provides the Predictor interface:

1334

#

1335

# ```py

1336

# class Predictor(object):

1337

# """Interface for constructing custom predictors."""

1338

#

1339

# def predict(self, instances, **kwargs):

1340

# """Performs custom prediction.

1341

#

1342

# Instances are the decoded values from the request. They have already

1343

# been deserialized from JSON.

1344

#

1345

# Args:

1346

# instances: A list of prediction input instances.

1347

# **kwargs: A dictionary of keyword args provided as additional

1348

# fields on the predict request body.

1349

#

1350

# Returns:

1351

# A list of outputs containing the prediction results. This list must

1352

# be JSON serializable.

1353

# """

1354

# raise NotImplementedError()

1355

#

1356

# @classmethod

1357

# def from_path(cls, model_dir):

1358

# """Creates an instance of Predictor using the given path.

1359

#

1360

# Loading of the predictor should be done in this method.

1361

#

1362

# Args:

1363

# model_dir: The local directory that contains the exported model

1364

# file along with any additional files uploaded when creating the

# version resource.

#

# Returns:

# An instance implementing this Predictor class.

1369

# """

1370

# raise NotImplementedError()

1371

# ```

1372

#

1373

# Learn more about [the Predictor interface and custom prediction

1374

# routines](/ml-engine/docs/tensorflow/custom-prediction-routines).

1375

"autoScaling": { # Options for automatically scaling a model. # Automatically scale the number of nodes used to serve the model in

1376

# response to increases and decreases in traffic. Care should be

1377

# taken to ramp up traffic according to the model's ability to scale

1378

# or you will start seeing increases in latency and 429 response codes.

1379

"minNodes": 42, # Optional. The minimum number of nodes to allocate for this model. These

1380

# nodes are always up, starting from the time the model is deployed.

1381

# Therefore, the cost of operating this model will be at least

1382

# `rate` * `min_nodes` * number of hours since last billing cycle,

1383

# where `rate` is the cost per node-hour as documented in the

1384

# [pricing guide](/ml-engine/docs/pricing),

1385

# even if no predictions are performed. There is additional cost for each

1386

# prediction performed.

1387

#

1388

# Unlike manual scaling, if the load gets too heavy for the nodes

1389

# that are up, the service will automatically add nodes to handle the

1390

# increased load as well as scale back as traffic drops, always maintaining

1391

# at least `min_nodes`. You will be charged for the time in which additional

1392

# nodes are used.

1393

#

1394

# If not specified, `min_nodes` defaults to 0, in which case, when traffic

1395

# to a model stops (and after a cool-down period), nodes will be shut down

1396

# and no charges will be incurred until traffic to the model resumes.

1397

#

1398

# You can set `min_nodes` when creating the model version, and you can also

1399

# update `min_nodes` for an existing version:

# <pre>

# update_body.json:

# {

# 'autoScaling': {

# 'minNodes': 5

# }

# }

# </pre>

# HTTP request:

# <pre>

# PATCH

# https://ml.googleapis.com/v1/{name=projects/*/models/*/versions/*}?update_mask=autoScaling.minNodes

1412

# -d @./update_body.json

1413

# </pre>

1414

},

1415

"serviceAccount": "A String", # Optional. Specifies the service account for resource access control.

1416

"state": "A String", # Output only. The state of a version.

1417

"pythonVersion": "A String", # Optional. The version of Python used in prediction. If not set, the default

1418

# version is '2.7'. Python '3.5' is available when `runtime_version` is set

1419

# to '1.4' and above. Python '2.7' works with all supported runtime versions.

1420

"framework": "A String", # Optional. The machine learning framework AI Platform uses to train

1421

# this version of the model. Valid values are `TENSORFLOW`, `SCIKIT_LEARN`,

1422

# `XGBOOST`. If you do not specify a framework, AI Platform

1423

# will analyze files in the deployment_uri to determine a framework. If you

1424

# choose `SCIKIT_LEARN` or `XGBOOST`, you must also set the runtime version

1425

# of the model to 1.4 or greater.

1426

#

1427

# Do **not** specify a framework if you're deploying a [custom

1428

# prediction routine](/ml-engine/docs/tensorflow/custom-prediction-routines).

1429

"packageUris": [ # Optional. Cloud Storage paths (`gs://…`) of packages for [custom

1430

# prediction routines](/ml-engine/docs/tensorflow/custom-prediction-routines)

1431

# or [scikit-learn pipelines with custom

1432

# code](/ml-engine/docs/scikit/exporting-for-prediction#custom-pipeline-code).

1433

#

1434

# For a custom prediction routine, one of these packages must contain your

1435

# Predictor class (see

1436

# [`predictionClass`](#Version.FIELDS.prediction_class)). Additionally,

1437

# include any dependencies used by your Predictor or scikit-learn pipeline

1438

# uses that are not already included in your selected [runtime

1439

# version](/ml-engine/docs/tensorflow/runtime-version-list).

1440

#

1441

# If you specify this field, you must also set

1442

# [`runtimeVersion`](#Version.FIELDS.runtime_version) to 1.4 or greater.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1443

"A String",

1444

],

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1445

"etag": "A String", # `etag` is used for optimistic concurrency control as a way to help

1446

# prevent simultaneous updates of a model from overwriting each other.

1447

# It is strongly suggested that systems make use of the `etag` in the

1448

# read-modify-write cycle to perform model updates in order to avoid race

1449

# conditions: An `etag` is returned in the response to `GetVersion`, and

1450

# systems are expected to put that etag in the request to `UpdateVersion` to

1451

# ensure that their change will be applied to the model as intended.

1452

"lastUseTime": "A String", # Output only. The time the version was last used for prediction.

1453

"deploymentUri": "A String", # Required. The Cloud Storage location of the trained model used to

1454

# create the version. See the

1455

# [guide to model

1456

# deployment](/ml-engine/docs/tensorflow/deploying-models) for more

1457

# information.

1458

#

1459

# When passing Version to

1460

# [projects.models.versions.create](/ml-engine/reference/rest/v1/projects.models.versions/create)

1461

# the model service uses the specified location as the source of the model.

1462

# Once deployed, the model version is hosted by the prediction service, so

1463

# this location is useful only as a historical record.

1464

# The total number of model files can't exceed 1000.

1465

"createTime": "A String", # Output only. The time the version was created.

1466

"isDefault": True or False, # Output only. If true, this version will be used to handle prediction

1467

# requests that do not specify a version.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1468

#

1469

# You can change the default version by calling

Sai Cheemalapati

2017-03-24 15:06:46 -0700

[diff] [blame]

1470

# [projects.methods.versions.setDefault](/ml-engine/reference/rest/v1/projects.models.versions/setDefault).

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1471

"name": "A String", # Required.The name specified for the version when it was created.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1472

#

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1473

# The version name must be unique within the model it is created in.

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1474

},

Bu Sun Kim

2019-06-14 16:50:42 -0700

[diff] [blame]

1475

"onlinePredictionLogging": True or False, # Optional. If true, online prediction access logs are sent to StackDriver

1476

# Logging. These logs are like standard server access logs, containing

1477

# information like timestamp and latency for each request. Note that

1478

# [Stackdriver logs may incur a cost](/stackdriver/pricing), especially if

1479

# your project receives prediction requests at a high queries per second rate

1480

# (QPS). Estimate your costs before enabling this option.

1481

#

1482

# Default is false.

1483

"name": "A String", # Required. The name specified for the model when it was created.

1484

#

1485

# The model name must be unique within the project it is created in.

1486

},

Sai Cheemalapati

2017-03-13 12:12:03 -0400

[diff] [blame]

1487

],

Sai Cheemalapati