[Bug] ADetailer Conversion Detector Error

[drone540]

Git commit

stable-diffusion.cpp master-797-5ef4a75

Operating System & Version

Kubuntu 24.04

GGML backends

Vulkan

Command-line arguments used

sd-cli -M adetailer -m /DATA/AppData/llama-swap-unified/models/stablediffusion/sd15/DreamShaper_8_pruned.safetensors -i generated-image-1785243512971-0.png -o repaired.png -p "close-up portrait of wonder woman, face visible" --ad-model /DATA/AppData/llama-swap-unified/models/stablediffusion/adetailer/face_yolov8n.safetensors --extra-ad-args "confidence=0.3"

Steps to reproduce

I have tried and seemingly failed to use the script to convert the adetailer face_yolov8n.pt to a safetensors that works with the following docker script that I used for conversion:

docker run --rm -it \
  -v "$PWD:/work" \
  -w /work \
  python:3.12 \
  bash -c "
    [ -f face_yolov8n.pt ] || wget https://huggingface.co/Bingsu/adetailer/resolve/main/face_yolov8n.pt &&
    [ -f convert_yolov8_to_safetensors.py ] || wget https://raw.githubusercontent.com/leejet/stable-diffusion.cpp/refs/heads/master/scripts/convert_yolov8_to_safetensors.py &&
    [ -f face_yolov8n.safetensors ] || (
      pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu &&
      pip install --no-cache-dir ultralytics safetensors &&
      pip uninstall -y opencv-python &&
      pip install --no-cache-dir opencv-python-headless &&
      python convert_yolov8_to_safetensors.py face_yolov8n.pt face_yolov8n.safetensors
    )
  "

What you expected to happen

The conversion to work properly and sd-server and sd-cli used adetailer

What actually happened

It claims to have converted it:

WARNING ⚠️ user config directory '/root/.config/Ultralytics' is not writeable, using '/tmp/Ultralytics'. Set YOLO_CONFIG_DIR to override.
Creating new Ultralytics Settings v0.0.6 file ✅
View Ultralytics Settings with 'yolo settings' or at '/tmp/Ultralytics/settings.json'
Update Settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'. For help see https://docs.ultralytics.com/quickstart/#ultralytics-settings.
YOLOv8n summary (fused): 72 layers, 3,005,843 parameters, 0 gradients, 8.1 GFLOPs
wrote face_yolov8n.safetensors: 126 tensors

Logs / error messages / stack trace

ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Quadro T1000 with Max-Q Design (NVIDIA) | uma: 0 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 49152 | int dot: 0 | matrix cores: none
load_backend: loaded Vulkan backend from /opt/ai/stable-diffusion.cpp/libggml-vulkan.so
load_backend: loaded CPU backend from /opt/ai/stable-diffusion.cpp/libggml-cpu-haswell.so
[INFO ] common.cpp:2158 - set width x height to 512 x 512
[INFO ] stable-diffusion.cpp:717  - loading model from '/DATA/AppData/llama-swap-unified/models/stablediffusion/sd15/DreamShaper_8_pruned.safetensors'
[INFO ] model_loader.cpp:242  - load /DATA/AppData/llama-swap-unified/models/stablediffusion/sd15/DreamShaper_8_pruned.safetensors using safetensors format
[INFO ] stable-diffusion.cpp:868  - Version: SD 1.x
[INFO ] stable-diffusion.cpp:923  - Weight type stat:                      f16: 1130
[INFO ] stable-diffusion.cpp:924  - Conditioner weight type stat:          f16: 196
[INFO ] stable-diffusion.cpp:925  - Diffusion model weight type stat:      f16: 686
[INFO ] stable-diffusion.cpp:926  - VAE weight type stat:                  f16: 248
[INFO ] stable-diffusion.cpp:1442 - using VAE for encoding / decoding
[INFO ] auto_encoder_kl.hpp:527  - vae decoder: ch = 128
[INFO ] stable-diffusion.cpp:1666 - total params memory size = 2035.00MB (VRAM 2035.00MB, RAM 0.00MB): text_encoders 235.06MB(VRAM), diffusion_model 1640.25MB(VRAM), vae 159.68MB(VRAM), controlnet 0.00MB(N/A), extensions 0.00MB(N/A)
[INFO ] stable-diffusion.cpp:1767 - running in eps-prediction mode
[WARN ] model_loader.cpp:252  - unknown format /DATA/AppData/llama-swap-unified/models/stablediffusion/adetailer/face_yolov8n.safetensors
[ERROR] detailer.cpp:681  - failed to load ADetailer detector: '/DATA/AppData/llama-swap-unified/models/stablediffusion/adetailer/face_yolov8n.safetensors'
[ERROR] main.cpp:591  - new_adetailer_ctx failed

Additional context / environment details

I have also tried with ultralytics==8.3.216 which is supposedly the last version before something changed that originally broke it.

I am not sure how to get this to actually work.

Also it seems that only sd-cli seems to call the adetailer? I did not see any errors reported when I was trying to get it to work with llama-swap from sd-server. It just generated the image and claimed adetailer was used in the pnginfo even though it never ran.