[bug]: When RAM cache is too low, LoRAs are not applied correctly

[psychedelicious]

Is there an existing issue for this problem?

• [X] I have searched the existing issues

Operating system

Linux

GPU vendor

Nvidia (CUDA)

GPU model

No response

GPU VRAM

24

Version number

4.2.1

Browser

FF

Python dependencies

No response

What happened

When the RAM cache is set too low (presumably a value too low to store a LoRA), the LoRA may not be applied correctly.

The following examples have the same prompt and seed - only the RAM cache and LoRA settings are changed.

Base test case metadata:

{
    "generation_mode": "sdxl_txt2img", 
    "positive_prompt": "super cute tiger cub alienzkin", 
    "negative_prompt": "", 
    "width": 1024, 
    "height": 1024, 
    "seed": 3944440447, 
    "rand_device": "cpu", 
    "cfg_scale": 5.5, 
    "cfg_rescale_multiplier": 0, 
    "steps": 50, 
    "scheduler": "dpmpp_2m_sde_k", 
    "model": {
        "key" : "e790edfe-1614-48fa-9802-58b83c0159b7"                                   , 
        "hash": "random:6cab136d48faf77462cab64f1f810971f1a6b925c94f1c6890bf8ae748936177", 
        "name": "Juggernaut-XL-v9"                                                       , 
        "base": "sdxl"                                                                   , 
        "type": "main"                                                                     
    }, 
    "loras": [
        {
            "model": {
                "key" : "3c106f7a-cdbc-4445-b11c-3915ac5886ee"                                   , 
                "hash": "random:bbebed7dd0694eeb822c9b7fac85d01f49738cf59f0a1c2b7116a38f8409257f", 
                "name": "alienzkin-sdxl"                                                         , 
                "base": "sdxl"                                                                   , 
                "type": "lora"                                                                     
            }, 
            "weight": 0.75
        }
    ], 
    "positive_style_prompt": "super cute tiger cub alienzkin", 
    "negative_style_prompt": "", 
    "control_layers": {"layers": [], "version": 2}, 
    "app_version": "4.2.1"
}

LoRA disabled (any ram setting)

ram: 7.5, LoRA enabled

ram: 0.25, LoRA enabled

What you expected to happen

LoRAs still work when the RAM cache setting is low.

How to reproduce the problem

No response

Additional context

No response

Discord username

No response