Anima lora for turbo generation, implemented with DP-DMD and Continuous-Time Distillation paper.

This model is presented in the paper When does training on downscaled images yield the same gradients?.

Try new anima_turbo_4step_v3.safetensors

trained with 10k steps, 30h in single 5070ti using anima_lora

Reproducible by CFG=1.0, steps=4, er_sde sampler

License

The weights in this repository are a Derivative of Anima and are distributed under the CircleStone Labs Non-Commercial License (see NOTICE.md). Any rights to use them are granted directly by CircleStone Labs LLC under that license. The ONNX exports under onnx/ embed the Anima model itself (with the turbo LoRA merged) and are covered by the same license.

The model/weights may only be used for non-commercial purposes. Generated images (Outputs) may be used commercially.
Anima is itself a Derivative Model of NVIDIA Cosmos-Predict2-2B-Text2Image and is subject to the NVIDIA Open Model License Agreement insofar as it applies to Derivative Models.
For a commercial license, see the Anima model card.
This is not an official CircleStone Labs product and is not endorsed by CircleStone Labs.

Earlier revisions of this card listed Apache-2.0; that was incorrect and is superseded by the license above.

Built on NVIDIA Cosmos.