Skip to content

Task: Introduce cosine matching loss #2481

Description

@ankitpatnala

Describe the task. It can be a feature, documentation, etc.

Upon initial experiments, we found the latents of the consecutive tokens have a cosine distance of 0.75.
Hence, we added a cosine loss to the consecutive forecasting latents.

Upon initial experiments, we show latents are more active and froct is not degrading.

Hedgedoc URL, if you are keeping notes, plots, logs in hedgedoc.

No response

Area

  • datasets, data readers, data preparation and transfer
  • model
  • science
  • infrastructure and engineering
  • evaluation, export and visualization
  • documentation
  • performance

Metadata

Metadata

Assignees

No one assigned

    Labels

    modelRelated to model training or definition (not generic infra)

    Type

    No type

    Projects

    Status
    No status

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions