Layout predictor weights — attribution and terms
================================================

These files are NNEF conversions of the layout model from surya
(https://github.com/VikParuchuri/surya), by Vik Paruchuri / Datalab,
checkpoint `2025_02_18` as distributed via https://models.datalab.to.

LICENCE
-------
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
(CC-BY-NC-SA-4.0). The full text is in `LICENSE` beside this file.

This differs from the licence of the files in the repository root, which are
Apache-2.0. See the repository README.

WAIVER (granted by the upstream author, not by this repository)
---------------------------------------------------------------
surya's documentation states:

  "The weights for the models are licensed `cc-by-nc-sa-4.0`, but I will waive
   that for any organization under $5M USD in gross revenue in the most recent
   12-month period AND under $5M in lifetime VC/angel funding raised. You also
   must not be competitive with the Datalab API."

That waiver is the upstream author's to grant. Redistributing these weights does
not extend it, and does not alter the CC-BY-NC-SA-4.0 terms above. Organizations
exceeding those limits, or wanting to remove the conditions, should contact
Datalab (https://www.datalab.to) about commercial licensing.

MODIFICATIONS
-------------
  - Format conversion from PyTorch to NNEF for the tract inference runtime,
    split into three graphs: vision encoder, decoder prefill, decoder iterative.
  - No changes to model architecture, and no fine-tuning or further training.

Under the ShareAlike term, adaptations of these files must be distributed under
CC-BY-NC-SA-4.0 as well.

RELATED
-------
surya's *code* is licensed GPL-3.0-or-later, separately from these weights.
Software that links a port of it inherits that copyleft; software that merely
loads these weights does not.