Package com.gencior.triton.core

Class InferResult

java.lang.Object
com.gencior.triton.core.InferResult
public class InferResult extends Object
An object of InferResult holds the response of an inference request and provides methods to retrieve inference results.
Author:
sachachoumiloff

  • Constructor Details

    • InferResult

      public InferResult(GrpcService.ModelInferResponse result)
      Constructs an InferResult instance that wraps a ModelInferResponse from the Triton server.

      This constructor takes the protobuf response message returned by the Triton Inference Server and provides convenient methods to access and deserialize the inference output tensors.

      Parameters:
      result - the ModelInferResponse returned by the Triton Inference Server, must not be null
      Throws:
      NullPointerException - if result is null

  • Method Details

    • getRequestId

      public String getRequestId()

    • getModelName

      public String getModelName()

    • getModelVersion

      public String getModelVersion()

    • getOutputNames

      public List<String> getOutputNames()

    • asFloatArray

      public float[] asFloatArray(String name)

    • asDoubleArray

      public double[] asDoubleArray(String name)

    • asIntArray

      public int[] asIntArray(String name)

    • asLongArray

      public long[] asLongArray(String name)

    • asStringArray

      public String[] asStringArray(String name)

    • getOutputAsArray

      public Object getOutputAsArray(String name)
      Retrieves the output tensor data as a native Java array.

      This method deserializes the inference output tensor identified by the given name into the appropriate Java primitive array type based on the Triton datatype. The method first attempts to deserialize from raw binary content for efficiency; if not available, it falls back to the contents field.

      The returned array type depends on the tensor's datatype:

      • BOOL → boolean[]
      • INT8 → byte[]
      • INT16 → short[]
      • INT32 → int[]
      • INT64 → long[]
      • UINT8 → short[]
      • UINT16 → int[]
      • UINT32 → long[]
      • UINT64 → long[]
      • FP32 → float[]
      • FP64 → double[]
      • FP16, BF16 → float[]
      • BYTES → String[]

      Parameters:
      name - the name of the output tensor to retrieve; must match the tensor name returned by the model
      Returns:
      a native array containing the deserialized tensor data
      Throws:
      TritonDataNotFoundException - if no output tensor with the specified name is found in the response
      TritonDataTypeException - if the tensor's datatype is not supported or invalid
      TritonInferException - if the tensor is found but contains no data

    • getOutput

      public GrpcService.ModelInferResponse.InferOutputTensor getOutput(String name)
      Retrieves the InferOutputTensor protobuf message for a specific output by name.

      This method provides direct access to the raw protobuf message representing the output tensor. It allows fine-grained control over the tensor data and metadata, including access to the tensor shape, datatype, and both raw and structured content representations.

      For most use cases, getOutputAsArray(String) is more convenient as it automatically deserializes the tensor into a native Java array. Use this method when you need direct access to the protobuf message or need to manually handle the tensor data.

      Parameters:
      name - the name of the output tensor to retrieve
      Returns:
      the InferOutputTensor protobuf message with the specified name, or null if no tensor with that name exists in the response

    • getResponse

      public GrpcService.ModelInferResponse getResponse()
      Retrieves the complete ModelInferResponse protobuf message.

      This method provides access to the entire response message from the Triton Inference Server, including all output tensors, model metadata, and any additional response information. This is useful when you need to access multiple outputs or inspect the complete response structure.

      Returns:
      the underlying ModelInferResponse protobuf message; never null