Class Kernel

• Defined in File kernel.h

Class Documentation

class Kernel

The Kernel class is used to define a MindSpore Kernel.

Public Functions

inline Kernel(const std::vector<mindspore::MSTensor> &inputs, const std::vector<mindspore::MSTensor> &outputs, const schema::Primitive *primitive, const mindspore::Context *ctx)

Constructor.

Parameters

• inputs – [in] define the input tensors for kernel.

• outputs – [in] define the output tensors for kernel.

• primitive – [in] define the primitive of kernel generated by flatbuffers.

• ctx – [in] define the context for kernel.

virtual ~Kernel() = default

Destructor.

virtual int Prepare() = 0

prepare for executing kernel.

Returns

result code.

virtual int Execute() = 0

execute the kernel.

Returns

result code.

virtual int ReSize() = 0

resize the kernel input shape, memory need to refresh.

Returns

result code.

inline virtual void set_inputs(const std::vector<mindspore::MSTensor> &in_tensors)

set kernel’s input tensors.

Parameters

in_tensors – [in] define the input tensors.

inline virtual void set_input(mindspore::MSTensor in_tensor, int index)

set kernel’s input tensor.

Parameters

• in_tensor – [in] define the input tensor.

• index – [in] define the index of the input tensor.

inline virtual void set_outputs(const std::vector<mindspore::MSTensor> &out_tensors)

set kernel’s output tensors.

Parameters

out_tensors – [in] define the output tensors.

inline virtual void set_output(mindspore::MSTensor out_tensor, int index)

set kernel’s output tensor.

Parameters

• out_tensor – [in] define the output tensor.

• index – [in] define the index of the output tensor.

inline virtual const std::vector<mindspore::MSTensor> &inputs()

obtain kernel’s input tensors.

Returns

input tensors.

inline virtual const std::vector<mindspore::MSTensor> &outputs()

obtain kernel’s output tensors.

Returns

output tensors.

inline std::string name() const

obtain kernel’s name.

Returns

kernel’s name.

inline void set_name(const std::string &name)

set kernel’s name.

Parameters

name – [in] define the kernel’s name.

inline const mindspore::Context *context() const

obtain kernel’s context.

Returns

kernel’s context.

inline virtual schema::PrimitiveType type() const

obtain kernel’s type.

Returns

kernel’s type.

inline virtual schema::QuantType quant_type() const

obtain kernel’s quant type.

Returns

kernel’s quant type.

inline const schema::Primitive *primitive() const

obtain the primitive of kernel generated by flatbuffers.

Returns

the primitive of kernel generated by flatbuffers.

inline std::string GetAttr(const std::string &key) const

get kernel’s attribute.

Parameters

key – [in] define the kernel’s attribute key.

inline void SetConfig(const std::map<std::string, std::map<std::string, std::string>> *config)

set kernel’s config.

Parameters

config – [in] define the kernel’s config.

inline std::map<std::string, std::string> GetConfig(const std::string &section) const

set kernel’s config.

Parameters

section – [in] define the section of the kernel’s config.