Lumina Image 2.0 as apache-2.0 model is not compatible with faipl-1.0-sd
Hi all,
First, congratulations on the release!
I noticed that this finetuned model, which is based on Lumina Image 2.0 (released under Apache 2.0), is being distributed under the Fair AI Public License 1.0-SD (faipl-1.0-sd).
This seems to be a licensing conflict. Apache 2.0 explicitly forbids imposing additional restrictions beyond those in the original license, and faipl-1.0-sd adds several use-case restrictions (“Prohibited Uses”). This makes the two licenses incompatible for direct relicensing or distribution of derivatives.
This is especially concerning since this model is the first large-scale finetune of Lumina Image 2.0. Using a non-permissive license here could gatekeep the community from building on top of Lumina and undermine the permissive, open source ecosystem that Apache 2.0 is meant to support.
Could the maintainers clarify how Neta-lumina 1.0 can be released under faipl-1.0-sd when its base model is Apache 2.0?
As it stands, this appears to violate the terms of the Apache 2.0 license.
Thanks for clarifying.
Hi, we have not quite realized this confliction, will discuss as a team to resolve this.
In any case, the final license won't be more restricted than common-sensed comercial friendly licensing.
Thank you for the heads up!
Apache 2.0 does not require derivative works to be licensed in Apache 2.0, it just needs attribution/credits.
For example, if I fork an apache 2.0 program, change the code (even 1 character), I can re-release it in whatever license I want, as long as I attribute the original code. It is not a copyleft license. (please refer to Section 4 of Apache 2.0)
I am not sure how it works for AI models however.