| MOSS RGBA | |
| This project adapts the following upstream work: | |
| MOSS-Image — OpenMOSS Team | |
| https://github.com/OpenMOSS/MOSS-Image | |
| Base model: https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT | |
| Upstream model and code license: Apache License 2.0. | |
| FLUX.2 Autoencoder — Black Forest Labs | |
| https://github.com/black-forest-labs/flux2#flux2-autoencoder | |
| The FLUX.2 autoencoder is separately released under Apache License 2.0. | |
| The RGBA VAE changes its image input/output to four channels and merges the learned adapter. | |
| This project's code and adapted weights are distributed under GPLv3. | |
| Upstream attribution and license terms are retained. | |
| See LICENSE and LICENSE-APACHE-2.0. | |