app.py

import gradio as gr
import numpy as np
import random

import spaces  # [uncomment to use ZeroGPU]
from sid import SiDFluxPipeline, SiDSD3Pipeline, SiDSanaPipeline
import torch
import os

os.environ["HF_HUB_DISABLE_PROGRESS_BARS"] = "1"
os.environ["HF_HUB_DISABLE_TELEMETRY"] = "1"

device = "cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16  # you can switch to bfloat16 if your GPU supports it

# Single model for this demo
MODEL_REPO_ID = "YGu1998/SiD-DiT-SD3-medium"

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1024

# ---- CACHING STATE ----
CACHED_PIPE = None
CACHED_TIME_SCALE = None


def load_model( progress=None):
    """
    Load the model once and cache it in globals.
    Subsequent calls reuse the same pipeline.
    """
    global CACHED_PIPE, CACHED_TIME_SCALE

    # If already loaded, reuse
    if CACHED_PIPE is not None:
        if progress is not None:
            progress(0.3, desc="Reusing cached model...")
        return CACHED_PIPE, CACHED_TIME_SCALE

    if progress is not None:
        progress(0.1, desc=f"Loading model from {MODEL_REPO_ID}...")

    time_scale = 1000.0  # for SANA Rectified Flow / TrigFlow

    # Load pipeline (you had bfloat16 here; keep if you like)
    pipe = SiDSD3Pipeline.from_pretrained(MODEL_REPO_ID, torch_dtype=torch_dtype)
    pipe = pipe.to(device)

    CACHED_PIPE = pipe
    CACHED_TIME_SCALE = time_scale

    if progress is not None:
        progress(0.5, desc="Model loaded")

    return pipe, time_scale


@spaces.GPU  # [uncomment to use ZeroGPU]
def infer(
    prompt,
    seed,
    randomize_seed,
    width,
    height,
    num_inference_steps,
    model_repo_id,  # in practice always MODEL_REPO_ID
    progress=gr.Progress(track_tqdm=False),
):
    # Seed handling
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    generator = torch.Generator().manual_seed(seed)

    # Phase 1: model loading / reuse
    progress(0.0, desc="Preparing model...")
    pipe, time_scale = load_model( progress=progress)

    # Phase 2: inference
    progress(0.7, desc="Running inference...")
    image = pipe(
        prompt=prompt,
        guidance_scale=1,
        num_inference_steps=num_inference_steps,
        width=width,
        height=height,
        generator=generator,
        time_scale=time_scale,
    ).images[0]

    progress(1.0, desc="Done")

    # IMPORTANT: do NOT delete the pipe if you want caching
    # pipe.maybe_free_model_hooks()
    # del pipe
    # torch.cuda.empty_cache()

    return image, seed


examples = [
    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
    "An astronaut riding a green horse",
    "A delicious ceviche cheesecake slice",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# SiD-DiT SD3-medium demo")

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            run_button = gr.Button("Run", scale=0, variant="primary")

        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )

            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)

            with gr.Row():
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                )

            with gr.Row():
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=4,
                    maximum=4,
                    step=1,
                    value=4,
                    interactive=False,  # read-only
                )

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

    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[
            prompt,
            seed,
            randomize_seed,
            width,
            height,
            num_inference_steps,
        ],
        outputs=[result, seed],
    )

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