Skip to content

Should we define IDL for MLActivation options as union of individual activation options? #363

@zolkis

Description

@zolkis

Currently we have 12 activation functions: clamp, elu, hardSigmoid, hardSwish, leakyRelu, linear, relu, sigmoid, softmax, softplus, softsign, tanh.

Should we define MLActivation options like 1 or 2?

  1. Generic dictionary + specifics in spec prose / algorithmic steps.
typedef object MLActivationOptions;
  1. Explicit union type
typedef (MLClampOptions or MLLeakyReluOptions or ...) MLActivationOptions;

There was a discussion in #337 here.
Looks like the current preference is 2.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions