사전 학습된 모델을 사용한 일괄 텍스트 예측

사전 학습된 텍스트 생성 모델을 사용하여 일괄 텍스트 예측을 실행합니다.

코드 샘플

Go

이 샘플을 사용해 보기 전에 Vertex AI 빠른 시작: 클라이언트 라이브러리 사용의 Go 설정 안내를 따르세요. 자세한 내용은 Vertex AI Go API 참고 문서를 참조하세요.

Vertex AI에 인증하려면 애플리케이션 기본 사용자 인증 정보를 설정합니다. 자세한 내용은 로컬 개발 환경의 인증 설정을 참조하세요.

import (
	"context"
	"fmt"
	"io"

	aiplatform "cloud.google.com/go/aiplatform/apiv1"
	aiplatformpb "cloud.google.com/go/aiplatform/apiv1/aiplatformpb"
	"google.golang.org/api/option"
	"google.golang.org/protobuf/types/known/structpb"
)

// batchTextPredict perform batch text prediction using a pre-trained text generation model
func batchTextPredict(w io.Writer, projectID, location, name, outputURI string, inputURIs []string) error {
	// inputURI := []string{"gs://cloud-samples-data/batch/prompt_for_batch_text_predict.jsonl"}
	// outputURI: existing template path. Following formats are allowed:
	// 	- gs://BUCKET_NAME/DIRECTORY/
	// 	- bq://project_name.llm_dataset

	ctx := context.Background()
	apiEndpoint := fmt.Sprintf("%s-aiplatform.googleapis.com:443", location)
	// Pretrained text model
	model := "publishers/google/models/text-bison"
	parameters := map[string]interface{}{
		"temperature":     0.2,
		"maxOutputTokens": 200,
		"topP":            0.95,
		"topK":            40,
	}

	parametersValue, err := structpb.NewValue(parameters)
	if err != nil {
		fmt.Fprintf(w, "unable to convert parameters to Value: %v", err)
		return err
	}

	client, err := aiplatform.NewJobClient(ctx, option.WithEndpoint(apiEndpoint))
	if err != nil {
		return err
	}
	defer client.Close()

	req := &aiplatformpb.CreateBatchPredictionJobRequest{
		Parent: fmt.Sprintf("projects/%s/locations/%s", projectID, location),
		BatchPredictionJob: &aiplatformpb.BatchPredictionJob{
			DisplayName:     name,
			Model:           model,
			ModelParameters: parametersValue,
			InputConfig: &aiplatformpb.BatchPredictionJob_InputConfig{
				Source: &aiplatformpb.BatchPredictionJob_InputConfig_GcsSource{
					GcsSource: &aiplatformpb.GcsSource{
						Uris: inputURIs,
					},
				},
				// List of supported formarts: https://cloud.google.com/vertex-ai/docs/reference/rpc/google.cloud.aiplatform.v1#model
				InstancesFormat: "jsonl",
			},
			OutputConfig: &aiplatformpb.BatchPredictionJob_OutputConfig{
				Destination: &aiplatformpb.BatchPredictionJob_OutputConfig_GcsDestination{
					GcsDestination: &aiplatformpb.GcsDestination{
						OutputUriPrefix: outputURI,
					},
				},
				// List of supported formarts: https://cloud.google.com/vertex-ai/docs/reference/rpc/google.cloud.aiplatform.v1#model
				PredictionsFormat: "jsonl",
			},
		},
	}

	job, err := client.CreateBatchPredictionJob(ctx, req)
	if err != nil {
		return err
	}
	fmt.Fprint(w, job.GetDisplayName())

	return nil
}

Java

이 샘플을 사용해 보기 전에 Vertex AI 빠른 시작: 클라이언트 라이브러리 사용의 Java 설정 안내를 따르세요. 자세한 내용은 Vertex AI Java API 참고 문서를 참조하세요.

Vertex AI에 인증하려면 애플리케이션 기본 사용자 인증 정보를 설정합니다. 자세한 내용은 로컬 개발 환경의 인증 설정을 참조하세요.

import com.google.cloud.aiplatform.v1.BatchPredictionJob;
import com.google.cloud.aiplatform.v1.GcsDestination;
import com.google.cloud.aiplatform.v1.GcsSource;
import com.google.cloud.aiplatform.v1.JobServiceClient;
import com.google.cloud.aiplatform.v1.JobServiceSettings;
import com.google.gson.Gson;
import com.google.protobuf.InvalidProtocolBufferException;
import com.google.protobuf.Value;
import com.google.protobuf.util.JsonFormat;
import java.io.IOException;
import java.util.HashMap;
import java.util.Map;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeoutException;

public class BatchTextPredictionSample {

  public static void main(String[] args)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String project = "YOUR_PROJECT_ID";
    String location = "us-central1";
    // inputUri: URI of the input dataset.
    // Could be a BigQuery table or a Google Cloud Storage file.
    // E.g. "gs://[BUCKET]/[DATASET].jsonl" OR "bq://[PROJECT].[DATASET].[TABLE]"
    String inputUri = "gs://cloud-samples-data/batch/prompt_for_batch_text_predict.jsonl";
    // outputUri: URI where the output will be stored.
    // Could be a BigQuery table or a Google Cloud Storage file.
    // E.g. "gs://[BUCKET]/[OUTPUT].jsonl" OR "bq://[PROJECT].[DATASET].[TABLE]"
    String outputUri = "gs://YOUR_BUCKET/batch_text_predict_output";
    String textModel = "text-bison";

    batchTextPrediction(project, inputUri, outputUri, textModel, location);
  }

  // Perform batch text prediction using a pre-trained text generation model.
  // Example of using Google Cloud Storage bucket as the input and output data source
  static BatchPredictionJob batchTextPrediction(
      String projectId, String inputUri, String outputUri, String textModel, String location)
      throws IOException {
    BatchPredictionJob response;
    JobServiceSettings jobServiceSettings =  JobServiceSettings.newBuilder()
        .setEndpoint("us-central1-aiplatform.googleapis.com:443").build();
    String parent = String.format("projects/%s/locations/%s", projectId, location);
    String modelName = String.format(
        "projects/%s/locations/%s/publishers/google/models/%s", projectId, location, textModel);
    // Construct model parameters
    Map<String, String> modelParameters = new HashMap<>();
    modelParameters.put("maxOutputTokens", "200");
    modelParameters.put("temperature", "0.2");
    modelParameters.put("topP", "0.95");
    modelParameters.put("topK", "40");
    Value parameterValue = mapToValue(modelParameters);

    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests.
    try (JobServiceClient jobServiceClient = JobServiceClient.create(jobServiceSettings)) {

      BatchPredictionJob batchPredictionJob =
          BatchPredictionJob.newBuilder()
              .setDisplayName("my batch text prediction job " + System.currentTimeMillis())
              .setModel(modelName)
              .setInputConfig(
                  BatchPredictionJob.InputConfig.newBuilder()
                      .setGcsSource(GcsSource.newBuilder().addUris(inputUri).build())
                      .setInstancesFormat("jsonl")
                      .build())
              .setOutputConfig(
                  BatchPredictionJob.OutputConfig.newBuilder()
                      .setGcsDestination(GcsDestination.newBuilder()
                          .setOutputUriPrefix(outputUri).build())
                      .setPredictionsFormat("jsonl")
                      .build())
              .setModelParameters(parameterValue)
              .build();

      // Create the batch prediction job
      response =
          jobServiceClient.createBatchPredictionJob(parent, batchPredictionJob);

      System.out.format("response: %s\n", response);
      System.out.format("\tName: %s\n", response.getName());
    }
    return response;
  }

  private static Value mapToValue(Map<String, String> map) throws InvalidProtocolBufferException {
    Gson gson = new Gson();
    String json = gson.toJson(map);
    Value.Builder builder = Value.newBuilder();
    JsonFormat.parser().merge(json, builder);
    return builder.build();
  }
}

Node.js

이 샘플을 사용해 보기 전에 Vertex AI 빠른 시작: 클라이언트 라이브러리 사용의 Node.js 설정 안내를 따르세요. 자세한 내용은 Vertex AI Node.js API 참고 문서를 참조하세요.

Vertex AI에 인증하려면 애플리케이션 기본 사용자 인증 정보를 설정합니다. 자세한 내용은 로컬 개발 환경의 인증 설정을 참조하세요.

// Imports the aiplatform library
const aiplatformLib = require('@google-cloud/aiplatform');
const aiplatform = aiplatformLib.protos.google.cloud.aiplatform.v1;

/**
 * TODO(developer):  Uncomment/update these variables before running the sample.
 */
// projectId = 'YOUR_PROJECT_ID';

// Optional: URI of the input dataset.
// Could be a BigQuery table or a Google Cloud Storage file.
// E.g. "gs://[BUCKET]/[DATASET].jsonl" OR "bq://[PROJECT].[DATASET].[TABLE]"
// inputUri =
//   'gs://cloud-samples-data/batch/prompt_for_batch_text_predict.jsonl';

// Optional: URI where the output will be stored.
// Could be a BigQuery table or a Google Cloud Storage file.
// E.g. "gs://[BUCKET]/[OUTPUT].jsonl" OR "bq://[PROJECT].[DATASET].[TABLE]"
// outputUri = 'gs://batch-bucket-testing/batch_text_predict_output';

// The name of batch prediction job
// jobDisplayName = `Batch text prediction job: ${new Date().getMilliseconds()}`;

// The name of pre-trained model
const textModel = 'text-bison';
const location = 'us-central1';

// Construct your modelParameters
const parameters = {
  maxOutputTokens: '200',
  temperature: '0.2',
  topP: '0.95',
  topK: '40',
};
const parametersValue = aiplatformLib.helpers.toValue(parameters);
// Configure the parent resource
const parent = `projects/${projectId}/locations/${location}`;
const modelName = `projects/${projectId}/locations/${location}/publishers/google/models/${textModel}`;

// Specifies the location of the api endpoint
const clientOptions = {
  apiEndpoint: `${location}-aiplatform.googleapis.com`,
};

// Instantiates a client
const jobServiceClient = new aiplatformLib.JobServiceClient(clientOptions);

// Perform batch text prediction using a pre-trained text generation model.
// Example of using Google Cloud Storage bucket as the input and output data source
async function callBatchTextPredicton() {
  const gcsSource = new aiplatform.GcsSource({
    uris: [inputUri],
  });

  const inputConfig = new aiplatform.BatchPredictionJob.InputConfig({
    gcsSource,
    instancesFormat: 'jsonl',
  });

  const gcsDestination = new aiplatform.GcsDestination({
    outputUriPrefix: outputUri,
  });

  const outputConfig = new aiplatform.BatchPredictionJob.OutputConfig({
    gcsDestination,
    predictionsFormat: 'jsonl',
  });

  const batchPredictionJob = new aiplatform.BatchPredictionJob({
    displayName: jobDisplayName,
    model: modelName,
    inputConfig,
    outputConfig,
    modelParameters: parametersValue,
  });

  const request = {
    parent,
    batchPredictionJob,
  };

  // Create batch prediction job request
  const [response] = await jobServiceClient.createBatchPredictionJob(request);

  console.log('Raw response: ', JSON.stringify(response, null, 2));
}

await callBatchTextPredicton();

Python

이 샘플을 사용해 보기 전에 Vertex AI 빠른 시작: 클라이언트 라이브러리 사용의 Python 설정 안내를 따르세요. 자세한 내용은 Vertex AI Python API 참고 문서를 참조하세요.

Vertex AI에 인증하려면 애플리케이션 기본 사용자 인증 정보를 설정합니다. 자세한 내용은 로컬 개발 환경의 인증 설정을 참조하세요.

from vertexai.preview.language_models import TextGenerationModel

# Example of using Google Cloud Storage bucket as the input and output data source
# TODO (Developer): Replace the input_uri and output_uri with your own GCS paths
# input_uri = "gs://cloud-samples-data/batch/prompt_for_batch_text_predict.jsonl"
# output_uri = "gs://your-bucket-name/batch_text_predict_output"

# Initialize the text generation model from a pre-trained model named "text-bison"
text_model = TextGenerationModel.from_pretrained("text-bison")

batch_prediction_job = text_model.batch_predict(
    dataset=input_uri,
    destination_uri_prefix=output_uri,
    # Optional:
    model_parameters={
        "maxOutputTokens": "200",
        "temperature": "0.2",
        "topP": "0.95",
        "topK": "40",
    },
)
print(batch_prediction_job.display_name)
print(batch_prediction_job.resource_name)
print(batch_prediction_job.state)

다음 단계

다른 Google Cloud 제품의 코드 샘플을 검색하고 필터링하려면 Google Cloud 샘플 브라우저를 참조하세요.