CC-BY-NC-4.0 License here vs MIT on GitHub
Could you please clarify how the cc-by-nc-4.0 license here and the MIT license on the BLASER page on Facebook Research’s GitHub correlate?
If I want to use the model for commercial purposes, what should I do?
The MIT license applies to the contents of the Github repository https://github.com/facebookresearch/stopes itself, i.e. the Python code in it, which includes the codebase for training or applying the BLASER models. The CC-BY-NC-4.0 license applies to the BLASER checkpoints (because they were trained using SONAR sentence encoder, which was itself licensed under CC-BY-NC-4.0, because it was trained using some datasets available only for non-commercial use). This means, the code could be used for any purpose (e.g. training and using some other models), but the particular set of BLASER 2.0 models can be used for non-commercial purposes only.
If Meta released the SONAR+BLASER models under a commercial-friendly license, it would violate the terms of use of the training dataset, which would create legal risks for it. If you use this model for commercial purposes, it could create legal risks for you. I am not in a position to assess the extent of these risks. And, unfortunately, not much could be done about it.
Thanks a lot, David. I'll close the thread.