Add torch compile option that can speed up models. Not tested with all models. Add support for edit datasets and Flux Kontext. See supported models doc for details. 2025-06-27 OmniGen2 LoRA training ...