Batch-Codevorhersage mit einem vortrainierten Modell

Batch-Codevorhersage mit einem vortrainierten Modell zur Codegenerierung durchführen.

Codebeispiel

Go

Bevor Sie dieses Beispiel anwenden, folgen Sie den Go-Einrichtungsschritten in der Vertex AI-Kurzanleitung zur Verwendung von Clientbibliotheken. Weitere Informationen finden Sie in der Referenzdokumentation zur Vertex AI Go API.

Richten Sie zur Authentifizierung bei Vertex AI Standardanmeldedaten für Anwendungen ein. Weitere Informationen finden Sie unter Authentifizierung für eine lokale Entwicklungsumgebung einrichten.

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"
)

// batchCodePredict perform batch code prediction using a pre-trained code generation model
func batchCodePredict(w io.Writer, projectID, location, name, outputURI string, inputURIs []string) error {
	// inputURI := []string{"gs://cloud-samples-data/batch/prompt_for_batch_code_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 code model
	model := "publishers/google/models/code-bison"
	parameters := map[string]interface{}{
		"temperature":     0.2,
		"maxOutputTokens": 200,
	}
	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

Bevor Sie dieses Beispiel anwenden, folgen Sie den Java-Einrichtungsschritten in der Vertex AI-Kurzanleitung zur Verwendung von Clientbibliotheken. Weitere Informationen finden Sie in der Referenzdokumentation zur Vertex AI Java API.

Richten Sie zur Authentifizierung bei Vertex AI Standardanmeldedaten für Anwendungen ein. Weitere Informationen finden Sie unter Authentifizierung für eine lokale Entwicklungsumgebung einrichten.

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.cloud.aiplatform.v1.LocationName;
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;

public class BatchCodePredictionSample {

  public static void main(String[] args) throws IOException, InterruptedException {
    // 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_code_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_code_predict_output";
    String codeModel = "code-bison";

    batchCodePredictionSample(project, location, inputUri, outputUri, codeModel);
  }

  // Perform batch code prediction using a pre-trained code generation model.
  // Example of using Google Cloud Storage bucket as the input and output data source
  public static BatchPredictionJob batchCodePredictionSample(
      String project, String location, String inputUri, String outputUri, String codeModel)
      throws IOException {
    BatchPredictionJob response;
    JobServiceSettings jobServiceSettings =  JobServiceSettings.newBuilder()
        .setEndpoint("us-central1-aiplatform.googleapis.com:443").build();
    LocationName parent = LocationName.of(project, location);
    String modelName = String.format(
        "projects/%s/locations/%s/publishers/google/models/%s", project, location, codeModel);
    // Construct your modelParameters
    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 client = JobServiceClient.create(jobServiceSettings)) {
      BatchPredictionJob batchPredictionJob =
          BatchPredictionJob.newBuilder()
              .setDisplayName("my batch code 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();

      response = client.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

Bevor Sie dieses Beispiel anwenden, folgen Sie den Node.js-Einrichtungsschritten in der Vertex AI-Kurzanleitung zur Verwendung von Clientbibliotheken. Weitere Informationen finden Sie in der Referenzdokumentation zur Vertex AI Node.js API.

Richten Sie zur Authentifizierung bei Vertex AI Standardanmeldedaten für Anwendungen ein. Weitere Informationen finden Sie unter Authentifizierung für eine lokale Entwicklungsumgebung einrichten.

// 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_code_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_code_predict_output';

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

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

// Construct your modelParameters
const parameters = {
  maxOutputTokens: '200',
  temperature: '0.2',
};
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/${codeModel}`;

// 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 code prediction using a pre-trained code generation model.
// Example of using Google Cloud Storage bucket as the input and output data source
async function callBatchCodePredicton() {
  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 callBatchCodePredicton();

Python

Bevor Sie dieses Beispiel anwenden, folgen Sie den Python-Einrichtungsschritten in der Vertex AI-Kurzanleitung zur Verwendung von Clientbibliotheken. Weitere Informationen finden Sie in der Referenzdokumentation zur Vertex AI Python API.

Richten Sie zur Authentifizierung bei Vertex AI Standardanmeldedaten für Anwendungen ein. Weitere Informationen finden Sie unter Authentifizierung für eine lokale Entwicklungsumgebung einrichten.

from vertexai.preview.language_models import CodeGenerationModel

# 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_code_predict.jsonl"
# output_uri = "gs://your-bucket-name/batch_code_predict_output"

code_model = CodeGenerationModel.from_pretrained("code-bison")

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

Nächste Schritte

Informationen zum Suchen und Filtern von Codebeispielen für andere Google Cloud -Produkte finden Sie im Google Cloud Beispielbrowser.