Skip to content

Why still use BatchNorm for spconv(embedding) and Pool/UnPool? #173

Description

@mikel-zhobro

Can you sharee the logic why Batchnorm is still used in the first spconv layer and in the (un)Pool layers?

Is there anything against replacing this with a LayerNorm, GroupNorm?

PS: If I understand it right, for the SerializedPooling, the BatchNorm would build the statistics over all points in all batches.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

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