app.py

import gradio as gr
import numpy as np
import random
import spaces
from PIL import Image

# import spaces #[uncomment to use ZeroGPU]
import torch

from transformers import AutoTokenizer, AutoModel
from models.gen_pipeline import NextStepPipeline
from utils.aspect_ratio import center_crop_arr_with_buckets

HF_HUB = "stepfun-ai/NextStep-1-Large"
device = "cuda" if torch.cuda.is_available() else "cpu"

tokenizer = AutoTokenizer.from_pretrained(HF_HUB, local_files_only=False, trust_remote_code=True)
model = AutoModel.from_pretrained(HF_HUB, local_files_only=False, trust_remote_code=True)
pipeline = NextStepPipeline(tokenizer=tokenizer, model=model).to(device=device)

MAX_SEED = np.iinfo(np.int16).max
MAX_IMAGE_SIZE = 512

DEFAULT_POSITIVE_PROMPT = None
DEFAULT_NEGATIVE_PROMPT = None

@spaces.GPU(duration=300)
def infer(
    prompt=None,
    #ref=None,
    seed=0,
    text_cfg=7.5,
    img_cfg=1.0,
    num_inference_steps=28,
    positive_prompt=DEFAULT_POSITIVE_PROMPT,
    negative_prompt=DEFAULT_NEGATIVE_PROMPT,
    progress=gr.Progress(track_tqdm=True),
):
    #if ref is None:
    #    gr.Warning("⚠️ Please upload an image!")
    #    return None

    if prompt in [None, ""]:
        gr.Warning("⚠️ Please enter a prompt!")
        return None
    IMG_SIZE = 512
    image = pipeline.generate_image(
        editing_caption,
        hw=(IMG_SIZE, IMG_SIZE),
        num_images_per_caption=1,
        positive_prompt=positive_prompt,
        negative_prompt=negative_prompt,
        cfg=text_cfg,
        cfg_img=img_cfg,
        cfg_schedule="constant",
        use_norm=False,
        num_sampling_steps=num_inference_steps,
        timesteps_shift=1.0,
        seed=seed,
        progress=True,
    )

    return image[0]

css = """
#col-container {
    margin: 0 auto;
    max-width: 800px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(" # NextStep-1-Large-Edit")

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )

            editing_caption = prompt

            run_button = gr.Button("Run", scale=0, variant="primary")

        with gr.Row():
            #ref = gr.Image(label="Reference Image", show_label=True, type="pil", height=400)

            with gr.Accordion("Advanced Settings", open=True):
                positive_prompt = gr.Text(
                    label="Positive Prompt",
                    show_label=False,
                    max_lines=1,
                    placeholder="Enter your positive prompt",
                    container=False,
                )
                negative_prompt = gr.Text(
                    label="Negative Prompt",
                    show_label=False,
                    max_lines=2,
                    placeholder="Enter your negative prompt",
                    container=False,
                )
                with gr.Row():
                    seed = gr.Slider(
                        label="Seed",
                        minimum=0,
                        maximum=MAX_SEED,
                        step=1,
                        value=42,
                    )
                    num_inference_steps = gr.Slider(
                        label="# sampling steps",
                        minimum=10,
                        maximum=50,
                        step=1,
                        value=28,  # Replace with defaults that work for your model
                    )

                with gr.Row():
                    text_cfg = gr.Slider(
                        label="Text cfg",
                        minimum=1.0,
                        maximum=15.0,
                        step=0.1,
                        value=7.5,  # Replace with defaults that work for your model
                    )
                    img_cfg = gr.Slider(
                        label="Image cfg",
                        minimum=1.0,
                        maximum=15.0,
                        step=0.1,
                        value=2.0,  # Replace with defaults that work for your model
                    )

        with gr.Row():
            result_1 = gr.Image(label="Result 1", show_label=False, container=True, height=400, visible=False)
            #result_2 = gr.Image(label="Result 2", show_label=False, container=True, height=400, visible=False)

        #gr.Examples(examples=examples, inputs=[prompt, ref])

    def show_result():
        return gr.update(visible=True)

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[
            prompt,
            #ref,
            seed,
            text_cfg,
            img_cfg,
            num_inference_steps,
            positive_prompt,
            negative_prompt,
        ],
        outputs=[result_1],
    )

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=show_result,
        outputs=[result_1],
    )


if __name__ == "__main__":
    demo.launch()