Skip to content

Fix input text chunking algorithm #28

@EndlessReform

Description

@EndlessReform

Currently,

  • Text chunks are too long (particularly in first token). This leads to long time-to-first-byte, dropped words, and pathological repetition
  • Final chunk is too short, leading to weird artifacts at the end of output

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions