README.md

---
tags:
- stable-diffusion-xl
- stable-diffusion-xl-diffusers
- diffusers-training
- text-to-image
- diffusers
- lora
- template:sd-lora

        
base_model: cookey39/aam_xl
instance_prompt: In the style of Terada,
license: openrail++
---

# SDXL LoRA DreamBooth - cookey39/teratera

<Gallery />

## Model description

## Generate Examples
https://www.pixiv.net/artworks/119150548

https://www.pixiv.net/artworks/119243202

https://www.pixiv.net/artworks/119243522

### These are cookey39/teratera LoRA adaption weights for cookey39/aam_xl.

## Download model

### Use it with UIs such as AUTOMATIC1111, Comfy UI, SD.Next, Invoke

- **LoRA**: download **[`teratera.safetensors` here 💾](/cookey39/teratera/blob/main/teratera.safetensors)**.
    - Place it on your `models/Lora` folder.
    - On AUTOMATIC1111, load the LoRA by adding `<lora:teratera:1>` to your prompt. On ComfyUI just [load it as a regular LoRA](https://comfyanonymous.github.io/ComfyUI_examples/lora/).
- *Embeddings*: download **[`teratera_emb.safetensors` here 💾](/cookey39/teratera/blob/main/teratera_emb.safetensors)**.
    - Place it on it on your `embeddings` folder
    - Use it by adding `teratera_emb` to your prompt. For example, `In the style of Terada,`
    (you need both the LoRA and the embeddings as they were trained together for this LoRA)
    

## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)

```py
from diffusers import AutoPipelineForText2Image
import torch
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
        
pipeline = AutoPipelineForText2Image.from_pretrained('stabilityai/stable-diffusion-xl-base-1.0', torch_dtype=torch.float16).to('cuda')
pipeline.load_lora_weights('cookey39/teratera', weight_name='pytorch_lora_weights.safetensors')
embedding_path = hf_hub_download(repo_id='cookey39/teratera', filename='teratera_emb.safetensors', repo_type="model")
state_dict = load_file(embedding_path)
# load embeddings of text_encoder 1 (CLIP ViT-L/14)
pipeline.load_textual_inversion(state_dict["clip_l"], token=["<s0>", "<s1>"], text_encoder=pipeline.text_encoder, tokenizer=pipeline.tokenizer)
# load embeddings of text_encoder 2 (CLIP ViT-G/14)
pipeline.load_textual_inversion(state_dict["clip_g"], token=["<s0>", "<s1>"], text_encoder=pipeline.text_encoder_2, tokenizer=pipeline.tokenizer_2)
        
instance_token = "<s0><s1>"
prompt = f"a {instance_token}masterpiece, best quality, full-length phoor portrait，Vibrant, solo, 1girl, smile, long hair, hair between eyes, multicolored eyes, hooded jacket, open jacket, shirt, long sleeves, ribbon, best quality, perfect anatomy, highres, absurdres{instance_token}  "
negative_prompt = "nsfw, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad feet"
image = pipeline(prompt=prompt, negative_prompt = negative_prompt, num_inference_steps=100, cross_attention_kwargs={"scale": 1.0},width = 960, height=1280).images[0]
image.save("./save.png")
```

For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)

## Trigger words

To trigger image generation of trained concept(or concepts) replace each concept identifier in you prompt with the new inserted tokens:

to trigger concept `TOK` → use `<s0><s1>` in your prompt 



## Details
All [Files & versions](/cookey39/teratera/tree/main).

The weights were trained using [🧨 diffusers Advanced Dreambooth Training Script](https://github.com/huggingface/diffusers/blob/main/examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py).

LoRA for the text encoder was enabled. False.

Pivotal tuning was enabled: True.

Special VAE used for training: None.