kpt-grpc-demo/google/cloud/aiplatform/v1/study.proto

// Copyright 2022 Google LLC
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

syntax = "proto3";

package google.cloud.aiplatform.v1;

import "google/api/field_behavior.proto";
import "google/api/resource.proto";
import "google/protobuf/duration.proto";
import "google/protobuf/struct.proto";
import "google/protobuf/timestamp.proto";

option csharp_namespace = "Google.Cloud.AIPlatform.V1";
option go_package = "google.golang.org/genproto/googleapis/cloud/aiplatform/v1;aiplatform";
option java_multiple_files = true;
option java_outer_classname = "StudyProto";
option java_package = "com.google.cloud.aiplatform.v1";
option php_namespace = "Google\\Cloud\\AIPlatform\\V1";
option ruby_package = "Google::Cloud::AIPlatform::V1";

// A message representing a Study.
message Study {
  option (google.api.resource) = {
    type: "aiplatform.googleapis.com/Study"
    pattern: "projects/{project}/locations/{location}/studies/{study}"
  };

  // Describes the Study state.
  enum State {
    // The study state is unspecified.
    STATE_UNSPECIFIED = 0;

    // The study is active.
    ACTIVE = 1;

    // The study is stopped due to an internal error.
    INACTIVE = 2;

    // The study is done when the service exhausts the parameter search space
    // or max_trial_count is reached.
    COMPLETED = 3;
  }

  // Output only. The name of a study. The study's globally unique identifier.
  // Format: `projects/{project}/locations/{location}/studies/{study}`
  string name = 1 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Required. Describes the Study, default value is empty string.
  string display_name = 2 [(google.api.field_behavior) = REQUIRED];

  // Required. Configuration of the Study.
  StudySpec study_spec = 3 [(google.api.field_behavior) = REQUIRED];

  // Output only. The detailed state of a Study.
  State state = 4 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. Time at which the study was created.
  google.protobuf.Timestamp create_time = 5 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. A human readable reason why the Study is inactive.
  // This should be empty if a study is ACTIVE or COMPLETED.
  string inactive_reason = 6 [(google.api.field_behavior) = OUTPUT_ONLY];
}

// A message representing a Trial. A Trial contains a unique set of Parameters
// that has been or will be evaluated, along with the objective metrics got by
// running the Trial.
message Trial {
  option (google.api.resource) = {
    type: "aiplatform.googleapis.com/Trial"
    pattern: "projects/{project}/locations/{location}/studies/{study}/trials/{trial}"
  };

  // A message representing a parameter to be tuned.
  message Parameter {
    // Output only. The ID of the parameter. The parameter should be defined in
    // [StudySpec's Parameters][google.cloud.aiplatform.v1.StudySpec.parameters].
    string parameter_id = 1 [(google.api.field_behavior) = OUTPUT_ONLY];

    // Output only. The value of the parameter.
    // `number_value` will be set if a parameter defined in StudySpec is
    // in type 'INTEGER', 'DOUBLE' or 'DISCRETE'.
    // `string_value` will be set if a parameter defined in StudySpec is
    // in type 'CATEGORICAL'.
    google.protobuf.Value value = 2 [(google.api.field_behavior) = OUTPUT_ONLY];
  }

  // Describes a Trial state.
  enum State {
    // The Trial state is unspecified.
    STATE_UNSPECIFIED = 0;

    // Indicates that a specific Trial has been requested, but it has not yet
    // been suggested by the service.
    REQUESTED = 1;

    // Indicates that the Trial has been suggested.
    ACTIVE = 2;

    // Indicates that the Trial should stop according to the service.
    STOPPING = 3;

    // Indicates that the Trial is completed successfully.
    SUCCEEDED = 4;

    // Indicates that the Trial should not be attempted again.
    // The service will set a Trial to INFEASIBLE when it's done but missing
    // the final_measurement.
    INFEASIBLE = 5;
  }

  // Output only. Resource name of the Trial assigned by the service.
  string name = 1 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The identifier of the Trial assigned by the service.
  string id = 2 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The detailed state of the Trial.
  State state = 3 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The parameters of the Trial.
  repeated Parameter parameters = 4 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The final measurement containing the objective value.
  Measurement final_measurement = 5 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. A list of measurements that are strictly lexicographically
  // ordered by their induced tuples (steps, elapsed_duration).
  // These are used for early stopping computations.
  repeated Measurement measurements = 6 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. Time when the Trial was started.
  google.protobuf.Timestamp start_time = 7 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. Time when the Trial's status changed to `SUCCEEDED` or `INFEASIBLE`.
  google.protobuf.Timestamp end_time = 8 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The identifier of the client that originally requested this Trial.
  // Each client is identified by a unique client_id. When a client
  // asks for a suggestion, Vertex AI Vizier will assign it a Trial. The client
  // should evaluate the Trial, complete it, and report back to Vertex AI
  // Vizier. If suggestion is asked again by same client_id before the Trial is
  // completed, the same Trial will be returned. Multiple clients with
  // different client_ids can ask for suggestions simultaneously, each of them
  // will get their own Trial.
  string client_id = 9 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. A human readable string describing why the Trial is
  // infeasible. This is set only if Trial state is `INFEASIBLE`.
  string infeasible_reason = 10 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The CustomJob name linked to the Trial.
  // It's set for a HyperparameterTuningJob's Trial.
  string custom_job = 11 [
    (google.api.field_behavior) = OUTPUT_ONLY,
    (google.api.resource_reference) = {
      type: "aiplatform.googleapis.com/CustomJob"
    }
  ];

  // Output only. URIs for accessing [interactive
  // shells](https://cloud.google.com/vertex-ai/docs/training/monitor-debug-interactive-shell)
  // (one URI for each training node). Only available if this trial is part of
  // a [HyperparameterTuningJob][google.cloud.aiplatform.v1.HyperparameterTuningJob] and the job's
  // [trial_job_spec.enable_web_access][google.cloud.aiplatform.v1.CustomJobSpec.enable_web_access] field
  // is `true`.
  //
  // The keys are names of each node used for the trial; for example,
  // `workerpool0-0` for the primary node, `workerpool1-0` for the first node in
  // the second worker pool, and `workerpool1-1` for the second node in the
  // second worker pool.
  //
  // The values are the URIs for each node's interactive shell.
  map<string, string> web_access_uris = 12 [(google.api.field_behavior) = OUTPUT_ONLY];
}

// Represents specification of a Study.
message StudySpec {
  // Represents a metric to optimize.
  message MetricSpec {
    // The available types of optimization goals.
    enum GoalType {
      // Goal Type will default to maximize.
      GOAL_TYPE_UNSPECIFIED = 0;

      // Maximize the goal metric.
      MAXIMIZE = 1;

      // Minimize the goal metric.
      MINIMIZE = 2;
    }

    // Required. The ID of the metric. Must not contain whitespaces and must be unique
    // amongst all MetricSpecs.
    string metric_id = 1 [(google.api.field_behavior) = REQUIRED];

    // Required. The optimization goal of the metric.
    GoalType goal = 2 [(google.api.field_behavior) = REQUIRED];
  }

  // Represents a single parameter to optimize.
  message ParameterSpec {
    // Value specification for a parameter in `DOUBLE` type.
    message DoubleValueSpec {
      // Required. Inclusive minimum value of the parameter.
      double min_value = 1 [(google.api.field_behavior) = REQUIRED];

      // Required. Inclusive maximum value of the parameter.
      double max_value = 2 [(google.api.field_behavior) = REQUIRED];

      // A default value for a `DOUBLE` parameter that is assumed to be a
      // relatively good starting point.  Unset value signals that there is no
      // offered starting point.
      //
      // Currently only supported by the Vertex AI Vizier service. Not supported
      // by HyperparameterTuningJob or TrainingPipeline.
      optional double default_value = 4;
    }

    // Value specification for a parameter in `INTEGER` type.
    message IntegerValueSpec {
      // Required. Inclusive minimum value of the parameter.
      int64 min_value = 1 [(google.api.field_behavior) = REQUIRED];

      // Required. Inclusive maximum value of the parameter.
      int64 max_value = 2 [(google.api.field_behavior) = REQUIRED];

      // A default value for an `INTEGER` parameter that is assumed to be a
      // relatively good starting point.  Unset value signals that there is no
      // offered starting point.
      //
      // Currently only supported by the Vertex AI Vizier service. Not supported
      // by HyperparameterTuningJob or TrainingPipeline.
      optional int64 default_value = 4;
    }

    // Value specification for a parameter in `CATEGORICAL` type.
    message CategoricalValueSpec {
      // Required. The list of possible categories.
      repeated string values = 1 [(google.api.field_behavior) = REQUIRED];

      // A default value for a `CATEGORICAL` parameter that is assumed to be a
      // relatively good starting point.  Unset value signals that there is no
      // offered starting point.
      //
      // Currently only supported by the Vertex AI Vizier service. Not supported
      // by HyperparameterTuningJob or TrainingPipeline.
      optional string default_value = 3;
    }

    // Value specification for a parameter in `DISCRETE` type.
    message DiscreteValueSpec {
      // Required. A list of possible values.
      // The list should be in increasing order and at least 1e-10 apart.
      // For instance, this parameter might have possible settings of 1.5, 2.5,
      // and 4.0. This list should not contain more than 1,000 values.
      repeated double values = 1 [(google.api.field_behavior) = REQUIRED];

      // A default value for a `DISCRETE` parameter that is assumed to be a
      // relatively good starting point.  Unset value signals that there is no
      // offered starting point.  It automatically rounds to the
      // nearest feasible discrete point.
      //
      // Currently only supported by the Vertex AI Vizier service. Not supported
      // by HyperparameterTuningJob or TrainingPipeline.
      optional double default_value = 3;
    }

    // Represents a parameter spec with condition from its parent parameter.
    message ConditionalParameterSpec {
      // Represents the spec to match discrete values from parent parameter.
      message DiscreteValueCondition {
        // Required. Matches values of the parent parameter of 'DISCRETE' type.
        // All values must exist in `discrete_value_spec` of parent parameter.
        //
        // The Epsilon of the value matching is 1e-10.
        repeated double values = 1 [(google.api.field_behavior) = REQUIRED];
      }

      // Represents the spec to match integer values from parent parameter.
      message IntValueCondition {
        // Required. Matches values of the parent parameter of 'INTEGER' type.
        // All values must lie in `integer_value_spec` of parent parameter.
        repeated int64 values = 1 [(google.api.field_behavior) = REQUIRED];
      }

      // Represents the spec to match categorical values from parent parameter.
      message CategoricalValueCondition {
        // Required. Matches values of the parent parameter of 'CATEGORICAL' type.
        // All values must exist in `categorical_value_spec` of parent
        // parameter.
        repeated string values = 1 [(google.api.field_behavior) = REQUIRED];
      }

      // A set of parameter values from the parent ParameterSpec's feasible
      // space.
      oneof parent_value_condition {
        // The spec for matching values from a parent parameter of
        // `DISCRETE` type.
        DiscreteValueCondition parent_discrete_values = 2;

        // The spec for matching values from a parent parameter of `INTEGER`
        // type.
        IntValueCondition parent_int_values = 3;

        // The spec for matching values from a parent parameter of
        // `CATEGORICAL` type.
        CategoricalValueCondition parent_categorical_values = 4;
      }

      // Required. The spec for a conditional parameter.
      ParameterSpec parameter_spec = 1 [(google.api.field_behavior) = REQUIRED];
    }

    // The type of scaling that should be applied to this parameter.
    enum ScaleType {
      // By default, no scaling is applied.
      SCALE_TYPE_UNSPECIFIED = 0;

      // Scales the feasible space to (0, 1) linearly.
      UNIT_LINEAR_SCALE = 1;

      // Scales the feasible space logarithmically to (0, 1). The entire
      // feasible space must be strictly positive.
      UNIT_LOG_SCALE = 2;

      // Scales the feasible space "reverse" logarithmically to (0, 1). The
      // result is that values close to the top of the feasible space are spread
      // out more than points near the bottom. The entire feasible space must be
      // strictly positive.
      UNIT_REVERSE_LOG_SCALE = 3;
    }

    oneof parameter_value_spec {
      // The value spec for a 'DOUBLE' parameter.
      DoubleValueSpec double_value_spec = 2;

      // The value spec for an 'INTEGER' parameter.
      IntegerValueSpec integer_value_spec = 3;

      // The value spec for a 'CATEGORICAL' parameter.
      CategoricalValueSpec categorical_value_spec = 4;

      // The value spec for a 'DISCRETE' parameter.
      DiscreteValueSpec discrete_value_spec = 5;
    }

    // Required. The ID of the parameter. Must not contain whitespaces and must be unique
    // amongst all ParameterSpecs.
    string parameter_id = 1 [(google.api.field_behavior) = REQUIRED];

    // How the parameter should be scaled.
    // Leave unset for `CATEGORICAL` parameters.
    ScaleType scale_type = 6;

    // A conditional parameter node is active if the parameter's value matches
    // the conditional node's parent_value_condition.
    //
    // If two items in conditional_parameter_specs have the same name, they
    // must have disjoint parent_value_condition.
    repeated ConditionalParameterSpec conditional_parameter_specs = 10;
  }

  // The decay curve automated stopping rule builds a Gaussian Process
  // Regressor to predict the final objective value of a Trial based on the
  // already completed Trials and the intermediate measurements of the current
  // Trial. Early stopping is requested for the current Trial if there is very
  // low probability to exceed the optimal value found so far.
  message DecayCurveAutomatedStoppingSpec {
    // True if [Measurement.elapsed_duration][google.cloud.aiplatform.v1.Measurement.elapsed_duration] is used as the x-axis of each
    // Trials Decay Curve. Otherwise, [Measurement.step_count][google.cloud.aiplatform.v1.Measurement.step_count] will be used
    // as the x-axis.
    bool use_elapsed_duration = 1;
  }

  // The median automated stopping rule stops a pending Trial if the Trial's
  // best objective_value is strictly below the median 'performance' of all
  // completed Trials reported up to the Trial's last measurement.
  // Currently, 'performance' refers to the running average of the objective
  // values reported by the Trial in each measurement.
  message MedianAutomatedStoppingSpec {
    // True if median automated stopping rule applies on
    // [Measurement.elapsed_duration][google.cloud.aiplatform.v1.Measurement.elapsed_duration]. It means that elapsed_duration
    // field of latest measurement of current Trial is used to compute median
    // objective value for each completed Trials.
    bool use_elapsed_duration = 1;
  }

  // Configuration for ConvexAutomatedStoppingSpec.
  // When there are enough completed trials (configured by
  // min_measurement_count), for pending trials with enough measurements and
  // steps, the policy first computes an overestimate of the objective value at
  // max_num_steps according to the slope of the incomplete objective value
  // curve. No prediction can be made if the curve is completely flat. If the
  // overestimation is worse than the best objective value of the completed
  // trials, this pending trial will be early-stopped, but a last measurement
  // will be added to the pending trial with max_num_steps and predicted
  // objective value from the autoregression model.
  message ConvexAutomatedStoppingSpec {
    // Steps used in predicting the final objective for early stopped trials. In
    // general, it's set to be the same as the defined steps in training /
    // tuning. If not defined, it will learn it from the completed trials. When
    // use_steps is false, this field is set to the maximum elapsed seconds.
    int64 max_step_count = 1;

    // Minimum number of steps for a trial to complete. Trials which do not have
    // a measurement with step_count > min_step_count won't be considered for
    // early stopping. It's ok to set it to 0, and a trial can be early stopped
    // at any stage. By default, min_step_count is set to be one-tenth of the
    // max_step_count.
    // When use_elapsed_duration is true, this field is set to the minimum
    // elapsed seconds.
    int64 min_step_count = 2;

    // The minimal number of measurements in a Trial.  Early-stopping checks
    // will not trigger if less than min_measurement_count+1 completed trials or
    // pending trials with less than min_measurement_count measurements. If not
    // defined, the default value is 5.
    int64 min_measurement_count = 3;

    // The hyper-parameter name used in the tuning job that stands for learning
    // rate. Leave it blank if learning rate is not in a parameter in tuning.
    // The learning_rate is used to estimate the objective value of the ongoing
    // trial.
    string learning_rate_parameter_name = 4;

    // This bool determines whether or not the rule is applied based on
    // elapsed_secs or steps. If use_elapsed_duration==false, the early stopping
    // decision is made according to the predicted objective values according to
    // the target steps. If use_elapsed_duration==true, elapsed_secs is used
    // instead of steps. Also, in this case, the parameters max_num_steps and
    // min_num_steps are overloaded to contain max_elapsed_seconds and
    // min_elapsed_seconds.
    bool use_elapsed_duration = 5;
  }

  // The available search algorithms for the Study.
  enum Algorithm {
    // The default algorithm used by Vertex AI for [hyperparameter
    // tuning](https://cloud.google.com/vertex-ai/docs/training/hyperparameter-tuning-overview)
    // and [Vertex AI Vizier](https://cloud.google.com/vertex-ai/docs/vizier).
    ALGORITHM_UNSPECIFIED = 0;

    // Simple grid search within the feasible space. To use grid search,
    // all parameters must be `INTEGER`, `CATEGORICAL`, or `DISCRETE`.
    GRID_SEARCH = 2;

    // Simple random search within the feasible space.
    RANDOM_SEARCH = 3;
  }

  // Describes the noise level of the repeated observations.
  //
  // "Noisy" means that the repeated observations with the same Trial parameters
  // may lead to different metric evaluations.
  enum ObservationNoise {
    // The default noise level chosen by Vertex AI.
    OBSERVATION_NOISE_UNSPECIFIED = 0;

    // Vertex AI assumes that the objective function is (nearly)
    // perfectly reproducible, and will never repeat the same Trial
    // parameters.
    LOW = 1;

    // Vertex AI will estimate the amount of noise in metric
    // evaluations, it may repeat the same Trial parameters more than once.
    HIGH = 2;
  }

  // This indicates which measurement to use if/when the service automatically
  // selects the final measurement from previously reported intermediate
  // measurements. Choose this based on two considerations:
  //  A) Do you expect your measurements to monotonically improve?
  //     If so, choose LAST_MEASUREMENT. On the other hand, if you're in a
  //     situation where your system can "over-train" and you expect the
  //     performance to get better for a while but then start declining,
  //     choose BEST_MEASUREMENT.
  //  B) Are your measurements significantly noisy and/or irreproducible?
  //     If so, BEST_MEASUREMENT will tend to be over-optimistic, and it
  //     may be better to choose LAST_MEASUREMENT.
  //  If both or neither of (A) and (B) apply, it doesn't matter which
  //  selection type is chosen.
  enum MeasurementSelectionType {
    // Will be treated as LAST_MEASUREMENT.
    MEASUREMENT_SELECTION_TYPE_UNSPECIFIED = 0;

    // Use the last measurement reported.
    LAST_MEASUREMENT = 1;

    // Use the best measurement reported.
    BEST_MEASUREMENT = 2;
  }

  oneof automated_stopping_spec {
    // The automated early stopping spec using decay curve rule.
    DecayCurveAutomatedStoppingSpec decay_curve_stopping_spec = 4;

    // The automated early stopping spec using median rule.
    MedianAutomatedStoppingSpec median_automated_stopping_spec = 5;

    // The automated early stopping spec using convex stopping rule.
    ConvexAutomatedStoppingSpec convex_automated_stopping_spec = 9;
  }

  // Required. Metric specs for the Study.
  repeated MetricSpec metrics = 1 [(google.api.field_behavior) = REQUIRED];

  // Required. The set of parameters to tune.
  repeated ParameterSpec parameters = 2 [(google.api.field_behavior) = REQUIRED];

  // The search algorithm specified for the Study.
  Algorithm algorithm = 3;

  // The observation noise level of the study.
  // Currently only supported by the Vertex AI Vizier service. Not supported by
  // HyperparameterTuningJob or TrainingPipeline.
  ObservationNoise observation_noise = 6;

  // Describe which measurement selection type will be used
  MeasurementSelectionType measurement_selection_type = 7;
}

// A message representing a Measurement of a Trial. A Measurement contains
// the Metrics got by executing a Trial using suggested hyperparameter
// values.
message Measurement {
  // A message representing a metric in the measurement.
  message Metric {
    // Output only. The ID of the Metric. The Metric should be defined in
    // [StudySpec's Metrics][google.cloud.aiplatform.v1.StudySpec.metrics].
    string metric_id = 1 [(google.api.field_behavior) = OUTPUT_ONLY];

    // Output only. The value for this metric.
    double value = 2 [(google.api.field_behavior) = OUTPUT_ONLY];
  }

  // Output only. Time that the Trial has been running at the point of this Measurement.
  google.protobuf.Duration elapsed_duration = 1 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. The number of steps the machine learning model has been trained for.
  // Must be non-negative.
  int64 step_count = 2 [(google.api.field_behavior) = OUTPUT_ONLY];

  // Output only. A list of metrics got by evaluating the objective functions using suggested
  // Parameter values.
  repeated Metric metrics = 3 [(google.api.field_behavior) = OUTPUT_ONLY];
}