anima-mlx / LICENSE
xocialize's picture
Anima-MLX: bf16 + int4 components (Cosmos-Predict2-2B anime T2I), NC
4cc65ee verified
Raw History Blame Contribute Delete
1.41 kB
Anima-MLX — License & Attribution
=================================
These weights are a format conversion (PyTorch → Apple MLX) of third-party models.
They are redistributed under NON-COMMERCIAL terms. Personal and research use only.
NOT for commercial use.
1. Fine-tune weights — Anima (CircleStone Labs)
The Anima anime/illustration fine-tune is licensed by CircleStone Labs under a
Non-Commercial license. This conversion inherits those terms.
Source: https://huggingface.co/circlestone-labs/Anima
2. Base denoiser — NVIDIA Cosmos-Predict2-2B
"Built on NVIDIA Cosmos." The base denoiser (CosmosTransformer3DModel) is licensed
under the NVIDIA Cosmos Open Model License. You must comply with its terms,
including its attribution requirement.
https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/
3. Other components
Qwen3-0.6B text encoder (Alibaba, Qwen license) and the Qwen-Image / Wan 16-channel
VAE (Alibaba/Wan) retain their respective upstream licenses.
4. Port code
The MLX port CODE (model definitions, pipeline, tools) is MIT-licensed and lives at
https://github.com/xocialize/anima-mlx . The MIT grant covers the code, not the weights.
By downloading or using these weights you agree to the upstream Non-Commercial terms above.
If any upstream license is more restrictive than this notice, the upstream license governs.