Skip to content

Rename swish to silu to give appropriate credit #8100

Description

@TFUsers

The swish was originally coined the "SiLU" in https://arxiv.org/pdf/1606.08415.pdf and https://arxiv.org/abs/1702.03118 long before the swish paper. Renaming other peoples' exact same ideas is unacceptable and huggingface's naming convention implicitly erases the research and work of people outside of Google.
This request inspired by a discussion and a recent tensorflow issue, but this problem has been brought up every few months for the past few years. In light of recent efforts to make the ML community more equitable and fair, this is a no-brainer and long overdue.

Will this change the current api? How?
The API would replace the "swish" argument with the "silu" argument and deprecate the swish.
PyTorch 1.7 added the SiLU. Tensorflow added the SiLU and should be in the next version. Jax has already added the SiLU;
jax.nn.swish will eventually be deprecated and jax.nn.silu will be added and both of the aforementioned papers will be cited in the documentation.

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