DataTypeList

Function Usage

Defines the operator parameter data types. If the data types supported by an input/output can be used in combination with the data types and formats supported by all other inputs/outputs, this API can be used to define the data types.

When DataType is used to configure data types, these data types are in one-to-one mapping with the data types and formats of other inputs and outputs. As shown in the following example, if the data type of input x and y is DT_FLOAT16, the data type of the corresponding output z is also DT_FLOAT16, and the supported data format must be FORMAT_ND.

class AddCustom : public OpDef {
public:
    AddCustom(const char* name) : OpDef(name)
    {
        this->Input("x")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Input("y")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Output("z")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        ...
    }
};

If the data types supported by an input/output can be used in combination with the data types and formats supported by all other inputs/outputs, the DataType API must be written in the following format, indicating that when input x is of type DT_FLOAT16, all data type and format combinations of input y and input z are supported.

class XxxCustom : public OpDef {
public:
    XxxCustom(const char* name) : OpDef(name)
    {
        this->Input("x")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT16, ge::DT_FLOAT16})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Input("y")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Output("z")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        ...
    }
};

In this case, you can use DataTypeList to specify the data types. For example:

class XxxCustom : public OpDef {
public:
    XxxCustom(const char* name) : OpDef(name)
    {
        this->Input("x")
            .ParamType(REQUIRED)
            .DataTypeList({ge::DT_FLOAT16})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Input("y")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        this->Output("z")
            .ParamType(REQUIRED)
            .DataType({ge::DT_FLOAT16, ge::DT_FLOAT, ge::DT_INT32})
            .Format({ge::FORMAT_ND, ge::FORMAT_ND, ge::FORMAT_ND});
        ...
    }
};

Prototype

OpParamDef &DataTypeList(std::vector<ge::DataType> types);

Parameters

Returns

OpParamDef operator definition. For details, see OpParamDef.

Constraints

• DataType and DataTypeList cannot be set for the same input or output at the same time.
• This API cannot be used together with UnknownShapeFormat.

Example

class AddCustom : public OpDef {
public:
    AddCustom(const char* name) : OpDef(name)
    {
        this->Input("x")
            .ParamType(REQUIRED)
            .DataTypeList({ge::DT_FLOAT})
            .Format({ge::FORMAT_ND, ge::FORMAT_NCHW});
        this->Input("x1")
             ......
    }
};