---
license: other
base_model: "Qwen/Qwen-Image"
tags:
  - qwen_image
  - qwen_image-diffusers
  - text-to-image
  - image-to-image
  - diffusers
  - simpletuner
  - safe-for-work
  - lora

  - template:sd-lora
  - lycoris
pipeline_tag: text-to-image
inference: true
widget:
- text: 'Graffito Mixed-Media Stop-Motion — MEDIUM SHOT of Tony, a puppet with a PHOTOGRAPHIC CUTOUT face and PAINTED PAPER CUTOUT body, running energetically towards the camera in a dimly lit street scene at night. Tony wears a light blue shirt with a red collar, his arms are outstretched, and his PHOTOGRAPHIC CUTOUT face shows an urgent expression. Behind him, a PAINTED CARDBOARD vehicle features a single bright headlight casting a harsh glare and illuminating Tony.'
  parameters:
    negative_prompt: 'blurry, cropped, ugly'
  output:
    url: ./assets/image_0_0.png
---

# gr4f1tt0_v1_qwen

This is a LyCORIS adapter derived from [Qwen/Qwen-Image](https://huggingface.co/Qwen/Qwen-Image).

The main validation prompt used during training was:
```
Graffito Mixed-Media Stop-Motion — MEDIUM SHOT of Tony, a puppet with a PHOTOGRAPHIC CUTOUT face and PAINTED PAPER CUTOUT body, running energetically towards the camera in a dimly lit street scene at night. Tony wears a light blue shirt with a red collar, his arms are outstretched, and his PHOTOGRAPHIC CUTOUT face shows an urgent expression. Behind him, a PAINTED CARDBOARD vehicle features a single bright headlight casting a harsh glare and illuminating Tony.
```


## Validation settings
- CFG: `3.5`
- CFG Rescale: `0.0`
- Steps: `30`
- Sampler: `FlowMatchEulerDiscreteScheduler`
- Seed: `None`
- Resolution: `1536x640`


Note: The validation settings are not necessarily the same as the [training settings](#training-settings).

You can find some example images in the following gallery:


<Gallery />

The text encoder **was not** trained.
You may reuse the base model text encoder for inference.


## Training settings

- Training epochs: 279
- Training steps: 7750
- Learning rate: 0.0001
  - Learning rate schedule: constant_with_warmup
  - Warmup steps: 100
- Max grad value: 0.01
- Effective batch size: 4
  - Micro-batch size: 1
  - Gradient accumulation steps: 4
  - Number of GPUs: 1
- Gradient checkpointing: False
- Prediction type: flow_matching[]
- Optimizer: adamw_bf16
- Trainable parameter precision: Pure BF16
- Base model precision: `fp8-torchao`
- Caption dropout probability: 0.0%



### LyCORIS Config:
```json
{
    "algo": "lokr",
    "multiplier": 1.0,
    "linear_dim": 10000,
    "linear_alpha": 1,
    "factor": 16,
    "apply_preset": {
        "target_module": [
            "Attention",
            "FeedForward"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 16
            },
            "FeedForward": {
                "factor": 8
            }
        }
    }
}
```


## Datasets

### graffito
- Repeats: 0
- Total number of images: 111
- Total number of aspect buckets: 1
- Resolution: 1.048576 megapixels
- Cropped: False
- Crop style: None
- Crop aspect: None
- Used for regularisation data: No


## Inference


```python
import torch
from diffusers import DiffusionPipeline
from lycoris import create_lycoris_from_weights


def download_adapter(repo_id: str):
    import os
    from huggingface_hub import hf_hub_download
    adapter_filename = "pytorch_lora_weights.safetensors"
    cache_dir = os.environ.get('HF_PATH', os.path.expanduser('~/.cache/huggingface/hub/models'))
    cleaned_adapter_path = repo_id.replace("/", "_").replace("\\", "_").replace(":", "_")
    path_to_adapter = os.path.join(cache_dir, cleaned_adapter_path)
    path_to_adapter_file = os.path.join(path_to_adapter, adapter_filename)
    os.makedirs(path_to_adapter, exist_ok=True)
    hf_hub_download(
        repo_id=repo_id, filename=adapter_filename, local_dir=path_to_adapter
    )

    return path_to_adapter_file
    
model_id = 'Qwen/Qwen-Image'
adapter_repo_id = 'davidrd123/gr4f1tt0_v1_qwen'
adapter_filename = 'pytorch_lora_weights.safetensors'
adapter_file_path = download_adapter(repo_id=adapter_repo_id)
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
lora_scale = 1.0
wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_file_path, pipeline.transformer)
wrapper.merge_to()

prompt = "Graffito Mixed-Media Stop-Motion — MEDIUM SHOT of Tony, a puppet with a PHOTOGRAPHIC CUTOUT face and PAINTED PAPER CUTOUT body, running energetically towards the camera in a dimly lit street scene at night. Tony wears a light blue shirt with a red collar, his arms are outstretched, and his PHOTOGRAPHIC CUTOUT face shows an urgent expression. Behind him, a PAINTED CARDBOARD vehicle features a single bright headlight casting a harsh glare and illuminating Tony."
negative_prompt = 'blurry, cropped, ugly'

## Optional: quantise the model to save on vram.
## Note: The model was quantised during training, and so it is recommended to do the same during inference time.
from optimum.quanto import quantize, freeze, qint8
quantize(pipeline.transformer, weights=qint8)
freeze(pipeline.transformer)
    
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu') # the pipeline is already in its target precision level
model_output = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(42),
    width=1536,
    height=640,
    guidance_scale=3.5,
).images[0]

model_output.save("output.png", format="PNG")

```