ai.djl.mxnet.engine

Class CachedOp

java.lang.Object

ai.djl.mxnet.jna.NativeResource

ai.djl.mxnet.engine.CachedOp

All Implemented Interfaces:

java.lang.AutoCloseable

public class CachedOp
extends NativeResource

The CachedOp is an internal helper that provides the core functionality to execute a MxSymbolBlock.

We don't recommended users interact with this class directly. Users should use Predictor instead. CachedOp is an operator that simplifies calling and analyzing the input shape. It requires minimum input to do inference because most of the information can be obtained from the model itself.

Field Summary

Fields inherited from class ai.djl.mxnet.jna.NativeResource

handle

Constructor Summary

Constructors

Constructor and Description
CachedOp(com.sun.jna.Pointer handle, MxNDManager manager, java.util.List<ai.djl.nn.Parameter> parameters, java.util.List<java.lang.Integer> paramIndices, ai.djl.util.PairList<java.lang.String,java.lang.Integer> dataIndices) Creates an instance of CachedOp.

Method Summary

Modifier and Type	Method and Description
void	close()
ai.djl.ndarray.NDList	forward(ai.djl.training.ParameterStore parameterStore, ai.djl.ndarray.NDList data) Assigns inputs to the empty locations of the input NDArray.

Methods inherited from class ai.djl.mxnet.jna.NativeResource

finalize, getHandle, getUid, isReleased

Methods inherited from class java.lang.Object

clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

CachedOp

public CachedOp(com.sun.jna.Pointer handle,
                MxNDManager manager,
                java.util.List<ai.djl.nn.Parameter> parameters,
                java.util.List<java.lang.Integer> paramIndices,
                ai.djl.util.PairList<java.lang.String,java.lang.Integer> dataIndices)

Creates an instance of CachedOp.

It can be created by using JnaUtils.createCachedOp(MxSymbolBlock, MxNDManager)

Parameters:

handle - the C handle of the CachedOp

manager - the manager used to create the NDArray

parameters - the parameter values

paramIndices - the parameters required by the model and their corresponding location

dataIndices - the input data names required by the model and their corresponding location

Method Detail

forward

public ai.djl.ndarray.NDList forward(ai.djl.training.ParameterStore parameterStore,
                                     ai.djl.ndarray.NDList data)

Assigns inputs to the empty locations of the input NDArray.

Parameters:

parameterStore - the parameterStore

data - the input in NDList format

Returns:

an NDList

close

public void close()

Specified by:

close in interface java.lang.AutoCloseable

Overrides:

close in class NativeResource