Introducing Google AI Edge Portal: Benchmark Edge AI at scale. Sign-up to request access during private preview.

litert::CustomOpKernel

This is an abstract class.

#include <litert_custom_op_kernel.h>

Defines the C++ interface for custom operator kernels in LiteRT.

Summary

Constructors and Destructors
`~CustomOpKernel()`

Public functions
`Destroy()=0`	`virtual Expected< void >`
`GetLiteRtCustomOpKernel()`	`LiteRtCustomOpKernel &`
`GetOutputLayouts(const std::vector< Layout > & input_layouts, std::vector< Layout > & output_layouts)=0`	`virtual Expected< void >`
`Init(const void *init_data, size_t init_data_size)=0`	`virtual Expected< void >`
`OpName() const =0`	`virtual const std::string &`
`OpVersion() const =0`	`virtual int`
`Run(const std::vector< TensorBuffer > & inputs, std::vector< TensorBuffer > & outputs)=0`	`virtual Expected< void >`

Protected functions
`CustomOpKernel()`

virtual Expected< void > Destroy()=0

LiteRtCustomOpKernel & GetLiteRtCustomOpKernel()

virtual Expected< void > GetOutputLayouts(
  const std::vector< Layout > & input_layouts,
  std::vector< Layout > & output_layouts
)=0

virtual Expected< void > Init(
  const void *init_data,
  size_t init_data_size
)=0

virtual const std::string & OpName() const =0

virtual int OpVersion() const =0

virtual Expected< void > Run(
  const std::vector< TensorBuffer > & inputs,
  std::vector< TensorBuffer > & outputs
)=0

virtual  ~CustomOpKernel()=default

 CustomOpKernel()