noob_v_pencil-XL / noob_v_pencil-XL-0.9r.md
bluepen5805's picture
Create noob_v_pencil-XL-0.9r.md
4018076 verified
|
raw
history blame
1.47 kB
# Tools
- [ComfyUI](https://github.com/comfyanonymous/ComfyUI)
- [sd-mecha](https://github.com/ljleb/sd-mecha)
- [sd-mecha for Comfyui](https://github.com/ljleb/comfy-mecha)
- [Concept Sliders](https://github.com/rohitgandikota/sliders)
# Checkpoints
- [NoobAI-XL (NAI-XL) V-Pred-0.9R-Version](https://civitai.com/models/833294?modelVersionId=1165792)
- [NoobAI-XL (NAI-XL) Epsilon-pred 1.0-version](https://civitai.com/models/833294?modelVersionId=1022833)
- [illustrious_pencil-XL-v2.0.0](https://huggingface.co/bluepen5805/illustrious_pencil-XL)
# Recipe
1. Merge using `sd-mecha for ComfyUI`. ( below sd-mecha format recipe )
```
model "sdxl\illustrious_pencil-XL-v2.0.0.safetensors" "sdxl" "base"
model "sdxl\noobaiXLNAIXL_vPred09RVersion.safetensors" "sdxl" "base"
model "sdxl\noobaiXLNAIXL_epsilonPred10Version.safetensors" "sdxl" "base"
merge "train_difference" &0 &1 &2 alpha=1.0
merge "clamped_add_opposite" &3 &0 &2 alpha=0.25
merge "clamped_add_opposite" &3 &1 &2 alpha=1.0
merge "add_cosine_a" &1 &5 alpha=0.75
merge "slerp" &4 &6 alpha=0.35000000000000003
```
2. Add `v_pred` and `ztsnr` keys.
```python
BASE_MODEL_NAME = "noob_v_pencil-XL-0.9r"
tensors = {}
with safe_open(BASE_MODEL_NAME + "-base.safetensors", framework="pt", device="cpu") as f:
for key in f.keys():
tensors[key] = f.get_tensor(key)
tensors["v_pred"] = torch.tensor([0.0])
tensors["ztsnr"] = torch.tensor([0.0])
save_file(tensors, BASE_MODEL_NAME + ".safetensors")
```