Builders & open source
Simo Ryu
Author of the LoRA repository for diffusion fine-tuning; AuraFlow collaborator and Diffusers contributor
Simo Ryu, who publishes on GitHub as cloneofsimo, wrote the Apache-2.0-licensed lora repository, which applies low-rank adaptation to quickly fine-tune text-to-image diffusion models, along with minimal reference implementations such as minDiffusion and minRF. Ryu collaborated with fal on AuraFlow v0.1, an Apache-2.0 open-weights flow-based text-to-image model whose model card credits Ryu with bringing the project to life. Ryu has also contributed to Hugging Face Diffusers, including a merged pull request adding learned positional-embedding selection for AuraFlow.123456
At a glance
- Current role
- None confirmed in the sources
- Work listed
- 3
- Sources cited
- 7
- Profile reviewed
Open-source work
Ryu's lora repository adapted low-rank adaptation to fine-tuning text-to-image diffusion models such as Stable Diffusion. Its update log, which runs from December 2022 to February 2023, records features including text-encoder fine-tuning, conversion to checkpoint formats used by other tools, pivotal tuning inversion, LoRA applied to ResNet layers, merging of LoRA weights and inpainting training.2
Ryu also writes minimal reimplementations of generative models, such as minDiffusion and minRF. In August 2023, Ryu contributed a pull request adding a minimalistic reimplementation of SDXL to the Diffusers documentation.16
AuraFlow
According to fal, Ryu set out to implement the MMDiT architecture from scratch in the minRF repository, and its early result, Lavenderflow-v0, led to a collaboration with fal to train a larger model. The result was AuraFlow v0.1, an open-weights flow-based text-to-image model released in July 2024 under the Apache 2.0 license.43
In August 2024, Ryu added learned positional-embedding selection for AuraFlow to Hugging Face Diffusers in a merged pull request.57
Work and contributions
Hugging Face Diffusers Open-source code
Authored merged pull requests to the Apache-2.0 Diffusers library, including #9182 (August 2024), which adds learned positional-embedding selection for AuraFlow, and #4532 (August 2023), which adds a section to the SDXL documentation pointing to minSDXL, a single-file implementation of the SDXL UNet.567
LoRA for diffusion fine-tuning (cloneofsimo/lora) Open-source code
Author of the repository "Using Low-rank adaptation to quickly fine-tune diffusion models", released under the Apache License 2.0; fal's AuraFlow announcement points to it as Ryu's adaptation of the LoRA paper for text-to-image models.214
AuraFlow v0.1 Open weights
Credited, with fal's @isidentical, on the AuraFlow model card for bringing the project to life; fal's announcement says the collaboration grew out of Ryu's minRF reimplementation of MMDiT. The model card lists the apache-2.0 license.34
Timeline
- Added text-encoder fine-tuning and checkpoint conversion to the lora repository.2
- Added pivotal tuning and adjustable ranks to the lora repository.2
- Added inpainting training with LoRA pivotal tuning to the lora repository.2
- Opened a Diffusers pull request adding a minimalistic reimplementation of SDXL to its documentation.6
- AuraFlow v0.1, built in collaboration with fal, was announced.4
- Opened a Diffusers pull request adding learned positional-embedding selection for AuraFlow.5
Documented roles
The sources do not confirm a current role; these are the roles they document.
Related people
- Hao ZhangImage, video, and 3D generation · Training systems
- Pieter NoordhuisTraining systems
- Hyung Won ChungTraining systems
- Danqi ChenTraining systems
- Vipul Ved PrakashTraining systems
- Byron HsuTraining systems
- Massimiliano Lupo PasiniTraining systems
- Fidelma M. RussoTraining systems
- Ron MokadyImage, video, and 3D generation
- Heiga ZenImage, video, and 3D generation
Suggest a correction or removal
Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.
References
7 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026
- cloneofsimo (Simo Ryu) on GitHub
- cloneofsimo/lora README and LICENSE (Apache-2.0)
- fal/AuraFlow model card (license apache-2.0)
- Introducing AuraFlow v0.1, an Open Exploration of Large Rectified Flow Models — fal blog
- huggingface/diffusers pull request #9182: Add Learned PE selection for Auraflow (merged)
- huggingface/diffusers pull request #4532: introduce minimalistic reimplementation of SDXL on the SDXL doc (merged)
- huggingface/diffusers LICENSE (Apache-2.0)