Show HN: UL-SMF – Open-source linear-complexity ~300x KV-cache compression

5 comments

• colingauvin 5 days ago ago

  This should be trivially demonstrable if it actually works. Cosine similarity is not a good metric, however, for attention compression.

  [-]
  • liventruth 4 days ago ago

    [flagged]

  • liventruth 4 days ago ago

    [flagged]

  • liventruth 4 days ago ago

    [dead]

• liventruth 5 days ago ago

  [flagged]