This model is trained using <a target="_blank" rel="ugc" href="https://modelscope.ai/civision/modelTraining">Civision/model training</a>. Training resources provided by ModelScope.<hr />V2: Training 10k steps, 295 images Cfg: 1.0"tatsten style. This is a black only image on white background of..."This is a complete retrain, went through the images again and refined a larger set. Including both the 9th &amp; 10th epochs, 10th goes a little more original than 9 but 9 is more consistently hitting a 1 in 4 (RNG dog witholds) gen of what's requested.Included edit (image to image) example of actual stencil output using this LoRA (9000). I'm sure the prompt can be refined, but even the base model without any LoRA is capable of creating a stencil that doesn't deviate far from the input reference with a good prompt.<hr />V1: Training: 8000 step, 100 images Cfg:1.0 "tattoo_stencil This is a digital black&amp;white image of..."Note: this run was about getting images that could be made into stencils, not about making the actual stencils. Give me a bit of time, I'm learning as I go; the goal IS to produce stencils either via image-to-image or directly.

Tattoo for Stencil Qwen21

V2 scratch retrain to 10,000 steps with 295 images - 9,000 step epoch produced best images.Example images generated with Qwen Image 2.1 Turbo using Viggle nodes setup (6 step instead of 4) on an RTX 4060 8GB.

tattoo_stencil_v2_c1-st9000.safetensors

tattoo_stencil_v2_c1-st10000.safetensors

v2.0

v1.0


# 

## Model Introduction

This model is trained using the [Civision/model training](https://modelscope.ai/civision/modelTraining) environment and computing resources provided by ModelScope.

* Model Type: LoRA
* Base Model: [Qwen/Qwen-Image-2.1](https://modelscope.ai/models/Qwen/Qwen-Image-2.1)
* Training Code: [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* Data amount：292
* Total training steps：10000
* License: Apache-2.0

## Inference Code

Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):

```bash
pip install diffsynth
```

Inference:

```python
from diffsynth.pipelines.qwen_image_21 import QwenImage21Pipeline, ModelConfig
import torch

pipe = QwenImage21Pipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="vae/diffusion_pytorch_model*.safetensors"),
    ],
    processor_config=ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="processor/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="joeblm/tattoo_stencil_v2", origin_file_pattern="tattoo_stencil_v2_c1-st10000.safetensors"))

prompt = "a cat"
image = pipe(prompt, seed=0, num_inference_steps=50)
image.save("image.png")
```


Files with SHA256: 384d5d5481b5bd94fc2ba9cf4839185664221d50c5c7ea543243a15256388244