Skip to main content

Collaboration Rules

To create a suitable environment in which every sector can come together and contribute effectively to the development of OpenThaiGPT, and to define shared goals effectively, we have established the following three collaboration rules.

  1. All work produced by the project must be released under the following licenses:
    1. Source Code / Weight / Model = Apache 2.0
    2. Dataset = CC BY-SA
  2. Access to resources (Open Resource) follows the information on the Open Resources page. Some resources are accessible only to the volunteer group, through the following process:
    1. Register as a member of a volunteer team.
    2. Begin contributing to the project in any way defined by the lead of that volunteer team, for example:
      1. Data Label Website team:
        1. Contribute at least 1 commit to the data tagging website.
      2. InstructDataset team:
        1. Tag at least 10 conversation pairs for the InstructDataset.
      3. RLHF team:
        1. Rank the model's generated outputs to build the reward model, for at least 10 conversation pairs.
      4. Pretraining team:
        1. Clean at least 10 articles of pretraining data.
        2. Take part in experiments to find a suitable LM architecture, for at least 1 configuration.
      5. OpenThaiGPT Library development team:
        1. Contribute at least 1 commit to the PIP OpenThaiGPT Library.
      6. Other contributions as appropriate.
    3. The volunteer team lead submits the names to the coordination team so that the website listing can be updated and access can be granted.
  3. If the OpenThaiGPT project name is used to apply for funding or for any other benefit, a letter of endorsement must be obtained from the OpenThaiGPT coordination team, issued jointly by the Artificial Intelligence Entrepreneur Association of Thailand (AIEAT) and the Artificial Intelligence Association of Thailand (AIAT) only.