Using Lora Causes OOM And Fills Ram

Nobody has claimed this yet.

### **Problem**
**(Im talking about Ram Not Video Card Memory/Vram)**

When using Any Lora in Normal SD Caused OOM Or my Ram getting filled up not unloading then i tried forge and its happening here as well.
Normal generation without Lora is fine and it dose not use 90% of my memory and it can generate and unload ram fine.
But as soon as i put Any Kinda of Lora on any Model it will fill 85% to 90% of my ram and crash on generating.

### **What did i try?**
I tried --medvram and --lowvram didn't solve it.
I tried Deleting Venv Folder and reinstalling didn't solve it.
i tried Reinstalling Whole forge and everything didn't solve it.
i tried reinstalling windows didn't solve it.

### **System Spec**
Ram: 16Gb ddr4
Gpu: Rtx 3060 12gb

### **webui-user.bat Settings**
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --no-download-sd-model --opt-split-attention --no-gradio-queue --skip-torch-cuda-test --opt-channelslast --autolaunch --ckpt-dir "D:\Ai\stable-webui\webui\models\Stable-diffusion" --lora-dir "D:\Ai\stable-webui\webui\models\Lora"
set CUDA_LAUNCH_BLOCKING=1
set SAFETENSORS_FAST_GPU=1

call webui.bat

### **Questions**
1.How do i fix it?

2.Are there any solutions for it like going back to a more stable version.

3.what version of forge or stable Dif you recommend me using?

Start with the reported webui-user.bat settings and reproduce generation with and without a LoRA on the stated 16 GB RAM and RTX 3060 system. Compare memory behavior and determine whether the issue can be isolated to LoRA loading; done means the RAM exhaustion is explained and a verified fix or clear reproducible finding is documented.

Written by the indexing model from the issue text.