File size: 1,500 Bytes
70ceb8b | 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 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 | # 🧩 ComfyUI-QwenVL — Custom Models Guide
You can add your own HuggingFace or local fine-tuned QwenVL models
without modifying the main code.
---
## 📁 File location
Place a file named `custom_models.json` in this folder:
```
ComfyUI/custom_nodes/ComfyUI-QwenVL/
````
When ComfyUI starts, it will automatically detect and merge this file
with the default `config.json`.
---
## ⚙️ File format
Use the following JSON structure:
```json
{
"hf_models": {
"My QwenVL Finetune": {
"repo_id": "myusername/MyQwenVL-Finetune",
"default": false,
"quantized": false,
"vram_requirement": {
"4bit": 4,
"8bit": 6,
"full": 10
}
},
"QwenVL Mini 4bit": {
"repo_id": "huggingface-user/qwenvl-mini-4bit",
"default": false,
"quantized": false,
"vram_requirement": {
"4bit": 2,
"8bit": 4,
"full": 8
}
}
}
}
````
* The **key** (e.g. `"My QwenVL Finetune"`) is what appears in the model dropdown list.
* `"repo_id"` can be a HuggingFace model name or a local directory path.
* `"vram_requirement"` is optional but helps ComfyUI display VRAM recommendations.
---
## 🧠 Notes
* If `custom_models.json` doesn’t exist, the system just skips it (no errors).
* You can edit and reload it anytime — simply restart ComfyUI to apply changes.
* Compatible with both Qwen2-VL and Qwen3-VL model architectures.
|