Wildcard

Versioned name: Wildcard-1

Category: Misc

Short description: Wildcard operation represents any compute logic and its input and output tensors contribute to the graph building.

Attributes

  • alias

    • Description: A list of strings of comma-separated indices so that the first index is the input index and the remaining indices are output indices aliasing this input. Take aten::relu_ as an example, the alias attribute would be: alias: [“0,0”] Here “0,0” means that input 0 (the first “0”) is aliased by output 0 (the second “0”).

    • Range of values: strings of comma-separated indices

    • Type: string[]

    • Default value: None

    • Required: no

  • mutation

    • Description: A list of input tensor indices that might be mutated by the Op. Take aten::relu_ as an example, the mutation attribute would be: Mutation: [0]

    • Range of values: Non-negative integers.

    • Type: int[]

    • Default value: None

    • Required: no

  • op_name

    • Description: indicate which OP Wildcard represents.

    • Range of values: string of op name

    • Type: string

    • Default value: None

    • Required: no

  • attr_names

    • Description: list of attributes’ name of represented OP

    • Range of values: strings of attributes’ names

    • Type: string[]

    • Default value: None

    • Required: no

  • attr_types

    • Description: list of attr_names’ date type. The order of attr_types should match attr_names’. For example, the first value in attr_types list indicates the data type of the first attribute in attr_names list.

    • Range of values: strings of data types:”i”,”is”,”f”,”fs”,”s”,”ss”,”b”,”bs”, which means “int”,”int[]”,”float”,”float[]”,”string”,”string[]”, “bool”,”bool[]”

    • Type: string[]

    • Default value: None

    • Required: no

Inputs:

  • 0 - N: input tensor. Optional.

Outputs

  • 0 - N: output tensor. Optional.