Introducing OmniLoRA: Train Any Model on Any GPU Inside Your ComfyUI Graph, supporting all backends, models and quantizations
retained text
Hello everybody! I've been working on this project for a while and I think it's finally ready to share. This node pack brings you an all in one universal lora trainer inside comfyUI. You get dataset preparation, training and monitoring in a single workflow. You can find it through the comfy manager or on github here: [https://github.com/cyberterm/ComfyUI-OmniLoRA](https://github.com/cyberterm/ComfyUI-OmniLoRA) The main goals of this project are: * Ease of use/installation with plug and play support and no extra dependencies. * Extensive model support with new models having a really high chance of being supported the moment comfyUI supports them without me interfering. * Support for every GPU + efficient CPU streaming to work on GPUs with less VRAM * GGUF support (you can easily train on q4, for example) * Modularity for powerful workflow integrations \*Supporting all models and GPUs is quite ambitious and it's impossible to test absolutely everything so please consider this a **beta version** and report any problems you have, I will get to them as soon as possible. You can also run the tests on your machine as described at the bottom of the readme and report any failed tests. Let me know what you think and if it works for you! Hopefully we can polish any rough edges together and build a workflow library for specific models with community fine-tuned parameters. Custom workflows can also support, for example, LLM dataset labeling or automatic subject separation and alpha masks for the dataset. Truly a lot of possibilities down the line!
document fetched 2026-11-18 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1xypn77/introducing_omnilora_train_any_model_on_any_gpu