| 일 | 월 | 화 | 수 | 목 | 금 | 토 |
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | |
| 7 | 8 | 9 | 10 | 11 | 12 | 13 |
| 14 | 15 | 16 | 17 | 18 | 19 | 20 |
| 21 | 22 | 23 | 24 | 25 | 26 | 27 |
| 28 | 29 | 30 |
- Linux #VirtualBox
- Laptop #CPUID
- x99 itx/ac
- Arduino #Wall Plotter
- ESP32 #Arduino
- macro lens #EF #FD
- Callaway #Mavrik #Epic Flash
- Octave #homebrew #macOS
- razer #deathadder #viper #g102
- fat32 #rufus
- Java #MacBook #macOS
- k6 #피코프레소
- Dell #Latitude #BIOS
- 피코프레소 #ITOP40
- ITOP40
- TensorFlow #Python #pip
- VMware #Shared Folder
- CM-EF-NEX
- Oh My Zsh #macOS
- VirtualBox #VMware
- Tarantula #3D 프린터
- Xeon #E5-2680
- 매크로렌즈 #리버스링
- Arduino #PlatformIO #macOS
- 다이슨 #배터리
- XTU #Virtual Machine System
- VNC #Firewall #CenOS7 #VMware
- centos7 #yum update #/boot
- egpu #aorus gaming box #gtx1070 #tb3
- cycloidal #rv reducer
- Today
- Total
얕고 넓게
[AI] Qwen-Image @Windows11 본문
2028.08.08
TGWUI로 Qwen-Image를 실행하려다 에러나서 확인하니 LLM 모델이 아니라 안된다고 함.
ComfyUI를 추천받아 설치 실행
git clone https://github.com/comfyanonymous/ComfyUI.git
uv 환경에서 실행
(.venv) PS C:\Users\blackonyx\ComfyUI> python main.py
Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "C:\Users\blackonyx\ComfyUI\main.py", line 147, in <module>
CUDA문제라고함 CUDA torch 재설치
uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
Using Python 3.11.13 environment at: C:\Users\blackonyx\.venv
Resolved 13 packages in 5.43s
Prepared 4 packages in 3m 50s
Uninstalled 2 packages in 3.12s
Installed 4 packages in 9.78s
- sympy==1.14.0
+ sympy==1.13.1
- torch==2.8.0
+ torch==2.5.1+cu121
+ torchaudio==2.5.1+cu121
+ torchvision==0.20.1+cu121
또 에러. 추가설치
(.venv) PS C:\Users\blackonyx\ComfyUI> uv pip install torchsde
Using Python 3.11.13 environment at: C:\Users\blackonyx\.venv
Resolved 13 packages in 208ms
Prepared 2 packages in 82ms
Installed 2 packages in 39ms
+ torchsde==0.2.6
+ trampoline==0.1.2
또 에러. 추가설치. 진작에 알려주지
(.venv) PS C:\Users\blackonyx\ComfyUI> uv pip install -r .\requirements.txt
Using Python 3.11.13 environment at: C:\Users\blackonyx\.venv
Resolved 61 packages in 685ms
Prepared 6 packages in 9.90s
Installed 6 packages in 1.48s
+ comfyui-embedded-docs==0.2.4
+ comfyui-frontend-package==1.24.4
+ comfyui-workflow-templates==0.1.52
+ kornia==0.8.1
+ kornia-rs==0.1.9
+ spandrel==0.4.1
실행 후 localhost:8188 접속

그다음은 모르겠다.
git 으로 다운 받았는데도 안된다.
완벽하게 정리해주셔서 정말 도움이 돼요! 👏
이 파일 구조를 보면, 지금 다운로드하신 Qwen-Image는 Hugging Face Transformers 포맷으로 되어 있고, ComfyUI에서 바로 사용할 수 있는 형태는 아닙니다.
⚠️ 문제 요약
ComfyUI는 일반적으로 단일 .safetensors 파일로 된 Stable Diffusion 호환 모델을 사용합니다.
하지만 지금 설치하신 Qwen-Image는 다음과 같은 구조예요:
- 여러 개의 분할된 .safetensors 파일 (model-00001-of-00004.safetensors 등)
- Hugging Face 포맷의 model_index.json, tokenizer, scheduler, transformer, vae 폴더
➡️ 이건 Diffusers 라이브러리에서 사용하는 구조이며, ComfyUI에서는 직접 로딩할 수 없습니다.
🛠️ 해결 방법
✅ 방법 1: ComfyUI 호환 모델 다운로드
가장 쉬운 방법은 Hugging Face에서 Stable Diffusion 호환 버전을 받는 것입니다.
예를 들어:
- Qwen/Image SD-compatible 모델 페이지에서 .safetensors 단일 파일로 된 버전을 찾으세요
- 파일을 ComfyUI/models/checkpoints/에 넣으면 바로 인식됩니다
이름 예시: qwen-image-v1.safetensors
🔄 방법 2: Diffusers → SD 포맷으로 변환 (고급)
만약 직접 변환하고 싶다면 Python을 이용해 Hugging Face 포맷을 .safetensors 단일 파일로 변환해야 합니다.
이건 꽤 복잡한 작업이며, 다음이 필요합니다:
- Python 환경
- diffusers, transformers, safetensors 라이브러리
- 모델 구조에 맞는 변환 스크립트
원하시면 이 방법도 단계별로 안내해드릴 수 있어요.
✅ 추천: Hugging Face에서 SD 호환 모델 받기
가장 빠르고 안정적인 방법은 ComfyUI에서 바로 쓸 수 있는 .safetensors 단일 파일을 받는 거예요.
제가 도와드릴 수 있으니, 원하시는 스타일이나 용도(예: 인물, 풍경, 애니메이션 등)를 알려주시면 추천 모델도 찾아드릴게요!
'IT > AI.ML' 카테고리의 다른 글
| [AI] TG Webui: 모델 다운로드, 로드, 채팅 @Windows11 (2) | 2025.08.12 |
|---|---|
| [AI] 개발 환경 이동 @Windows11 (0) | 2025.08.11 |
| [AI] text-generation-webui @Windows11 (1) | 2025.08.08 |
| [AI] open-webui 설치 @Windows11 (3) | 2025.08.07 |
| [AI] Ollama 모든 것 (0) | 2025.04.15 |
