File size: 2,195 Bytes
0ed6b0e | 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 | # Inference WebUI
DiffSynth-Studio provides an Inference WebUI to help developers quickly validate model performance.
> The current Inference WebUI is not fully developed yet; we will optimize the interaction logic in the future.
> The Inference WebUI is a debugging tool designed for developers, not a creation tool for end-users. For a richer feature set and more user-friendly interactive experience, we recommend using the [AIGC Zone](https://modelscope.cn/aigc/home) on ModelScope (for users in China) or the [Civision Zone](https://modelscope.ai/civision/home) (for users outside China).
## Launching the Inference WebUI
The Inference WebUI is built on [`Streamlit`](https://streamlit.io/). In addition to DiffSynth-Studio, you also need to install `Streamlit`:
```shell
git clone https://github.com/modelscope/DiffSynth-Studio.git
cd DiffSynth-Studio
pip install -e .
pip install streamlit
```
Launch command:
```shell
streamlit run examples/dev_tools/webui.py --server.fileWatcherType none
```
## How It Works
As a standalone tool, the Inference WebUI dynamically generates corresponding UI controls by parsing the type annotations of parameters in the Pipeline's `from_pretrained` and `__call__` methods. Therefore, the interface interaction logic is fully consistent with the code invocation logic, serving as a visual entry point for DiffSynth-Studio code.
Taking `ZImagePipeline.__call__` in `diffsynth.pipelines.z_image` as an example:
```python
@torch.no_grad()
def __call__(
self,
# Prompt
prompt: str = "",
negative_prompt: str = "",
cfg_scale: float = 1.0,
# Image
input_image: Image.Image = None,
denoising_strength: float = 1.0,
...
)
```
After parsing, the WebUI will automatically render the following interface:
## Usage Tips
- Supports automatic loading of model information such as `model_id` and `origin_file_pattern` from sample code in `./examples`, simplifying the configuration process;
- Parameters such as `vram_limit`, `tokenizer_config`, and `lora` cannot be retrieved through code parsing and need to be filled in manually.
|