License for the model

[Tatyana-jl]

Hi! May I kindly ask you to clarify the license for the model paraphrase-xlm-r-multilingual-v1? Is it licensed with Apache2-0 like others (it does not have any tags in HuggingFace library)?

[PhilipMay]

Should be the same as the base Model https://huggingface.co/xlm-roberta-base

So MIT license.

Am 19.04.2021 um 13:51 schrieb Tatyana-jl @.***>:

 Hi! May I kindly ask you to clarify the license for the model paraphrase-xlm-r-multilingual-v1? Is it licensed with Apache2-0 like others (it does not have any tags in HuggingFace library)?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[Tatyana-jl]

Thank you!

[PhilipMay]

Well - I would recommend to wait for an answer of @nreimers .

[nreimers]

Hi @Tatyana-jl licensing is a complex topic, sadly we cannot provide there a binding answer.

The questions are:

• The model is derived from xlm-robert-base, which has a MIT license => Does the derive model also must have MIT license?
• The model is a distillation of the English paraphrase model. The English paraphrase model was trained on various paraphrase datasets => does any of the datasets enforce a license on the English paraphrase model? => And does this enforce any license on the multilingual model?
• The distillation process was done with parallel data from various copora: Does any of the corpora force a specific license on the model via "derived work"?
• Also licensing can depend on the country, e.g. USA vs. Europe might be different.

From our side you can use the model with MIT or Apache license. But I don't know if there are other license restrictions that enforce a specific license.