Lora Weight Composition And Inference Time Model Merging

via “lora and weight adapter composition with dynamic weight merging”

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.

Unique: Dynamic LoRA composition with per-adapter strength multipliers and multi-LoRA stacking, enabling real-time weight blending without model retraining or disk I/O

vs others: More flexible than static LoRA merging because weights are blended at inference time; supports more LoRAs per workflow than WebUI's sequential loading