app.py

import gradio as gr
import numpy as np
import random

import spaces  # [uncomment to use ZeroGPU]
from diffusers import DiffusionPipeline
import torch
from PIL import Image

model_ids = [
    "Prgckwb/trpfrog-sd3.5-large",
    "Prgckwb/trpfrog-diffusion",
]

if torch.cuda.is_available():
    torch_dtype = torch.float16
    device = "cuda"
else:
    torch_dtype = torch.float32
    device = "cpu"

pipelines = {
    model_id: DiffusionPipeline.from_pretrained(
        model_id, torch_dtype=torch_dtype
    ) if device == 'cuda' else None
    for model_id in model_ids
}


@spaces.GPU()
@torch.inference_mode()
def inference(
    model_id: str,
    prompt: str,
    width: int,
    height: int,
    progress=gr.Progress(track_tqdm=True),
):
    if device == 'cuda':
        pipe = pipelines[model_id].to(device)

        image = pipe(
            prompt=prompt,
            width=width,
            height=height,
        ).images[0]
    else:
        # 真っ黒の画像を生成
        image = Image.fromarray(np.random.randn(height, width, 3).astype(np.uint8))

    return image

def create_interface():
    theme = gr.themes.Ocean()

    with gr.Blocks(theme=theme) as demo:
        with gr.Column():
            gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>TrpFrog Diffusion Demo</h1>")

            with gr.Row():
                with gr.Column():
                    input_model_id = gr.Dropdown(label="Model", choices=model_ids, value=model_ids[0])
                    input_prompt = gr.Textbox(label="Prompt", placeholder="an icon of trpfrog", value="an icon of trpfrog")

                    with gr.Row():
                        input_width = gr.Slider(label="Width", minimum=64, maximum=2056, step=128, value=1024)
                        input_height = gr.Slider(label="Height", minimum=64, maximum=2056, step=128, value=1024)

                    with gr.Row():
                        clear_btn = gr.ClearButton(components=[input_prompt])
                        submit_btn = gr.Button('Generate', variant='primary')

                with gr.Column():
                    output_image = gr.Image(label="Output")
            all_inputs = [input_model_id, input_prompt, input_width, input_height]
            all_outputs = [output_image]

            examples = gr.Examples(
                examples=[
                    ['Prgckwb/trpfrog-sd3.5-large', 'an icon of trpfrog eating ramen', 1024, 1024],
                    ['Prgckwb/trpfrog-sd3.5-large', 'an icon of trpfrog with a gun', 1024, 1024],
                ],
                inputs=all_inputs,
                outputs=all_outputs,
                fn=inference,
                cache_mode='eager',
                cache_examples=True,
            )

        submit_btn.click(inference, inputs=all_inputs, outputs=all_outputs)
        input_prompt.submit(inference, inputs=all_inputs, outputs=all_outputs)

    return demo

if __name__ == "__main__":
    try:
        demo = create_interface()
        demo.queue().launch()
    except Exception as e:
        raise gr.Error(e)