app.py

from typing import Any

import gradio as gr
import PIL
import spaces
import torch
from hi_diffusers import HiDreamImagePipeline, HiDreamImageTransformer2DModel
from hi_diffusers.schedulers.flash_flow_match import (
    FlashFlowMatchEulerDiscreteScheduler,
)
from transformers import LlamaForCausalLM, PreTrainedTokenizerFast

# Constants
MODEL_PREFIX: str = "HiDream-ai"
LLAMA_MODEL_NAME: str = "meta-llama/Meta-Llama-3.1-8B-Instruct"
MODEL_PATH = "HiDream-ai/HiDream-I1-Dev"
MODEL_CONFIGS: dict[str, Any] = {
    "guidance_scale": 0.0,
    "num_inference_steps": 28,
    "shift": 6.0,
    "scheduler": FlashFlowMatchEulerDiscreteScheduler,
}

# Model configurations
# MODEL_CONFIGS: dict[str, dict] = {
#     "full": {
#         "path": f"{MODEL_PREFIX}/HiDream-I1-Full",
#         "guidance_scale": 5.0,
#         "num_inference_steps": 50,
#         "shift": 3.0,
#         "scheduler": FlowUniPCMultistepScheduler,
#     },
#     "fast": {
#         "path": f"{MODEL_PREFIX}/HiDream-I1-Fast",
#         "guidance_scale": 0.0,
#         "num_inference_steps": 16,
#         "shift": 3.0,
#         "scheduler": FlashFlowMatchEulerDiscreteScheduler,
#     },
# }

# Supported image sizes
RESOLUTION_OPTIONS: list[str] = [
    "1024 x 1024 (Square)",
    "768 x 1360 (Portrait)",
    "1360 x 768 (Landscape)",
    "880 x 1168 (Portrait)",
    "1168 x 880 (Landscape)",
    "1248 x 832 (Landscape)",
    "832 x 1248 (Portrait)",
]


tokenizer = PreTrainedTokenizerFast.from_pretrained(LLAMA_MODEL_NAME, use_fast=False)
text_encoder = LlamaForCausalLM.from_pretrained(
    LLAMA_MODEL_NAME,
    output_hidden_states=True,
    output_attentions=True,
    torch_dtype=torch.bfloat16,
).to("cuda")

transformer = HiDreamImageTransformer2DModel.from_pretrained(
    MODEL_PATH,
    subfolder="transformer",
    torch_dtype=torch.bfloat16,
).to("cuda")

scheduler = MODEL_CONFIGS["scheduler"](
    num_train_timesteps=1000,
    shift=MODEL_CONFIGS["shift"],
    use_dynamic_shifting=False,
)

pipe = HiDreamImagePipeline.from_pretrained(
    MODEL_PATH,
    scheduler=scheduler,
    tokenizer_4=tokenizer,
    text_encoder_4=text_encoder,
    torch_dtype=torch.bfloat16,
).to("cuda", torch.bfloat16)

pipe.transformer = transformer


@spaces.GPU(duration=90)
def generate_image(
    prompt: str,
    resolution: str,
    seed: int,
) -> tuple[PIL.Image.Image, int]:
    if seed == -1:
        seed = torch.randint(0, 1_000_000, (1,)).item()

    height, width = tuple(map(int, resolution.replace(" ", "").split("x")))
    generator = torch.Generator("cuda").manual_seed(seed)

    image = pipe(
        prompt=prompt,
        height=height,
        width=width,
        guidance_scale=MODEL_CONFIGS["guidance_scale"],
        num_inference_steps=MODEL_CONFIGS["num_inference_steps"],
        generator=generator,
    ).images[0]

    torch.cuda.empty_cache()
    return image, seed


# Gradio UI
with gr.Blocks(title="HiDream Image Generator") as demo:
    gr.Markdown("## 🌈 HiDream Image Generator")

    with gr.Row():
        with gr.Column():
            prompt = gr.Textbox(
                label="Prompt",
                placeholder="e.g. A futuristic city with floating cars at sunset",
                lines=3,
            )

            resolution = gr.Radio(
                choices=RESOLUTION_OPTIONS,
                value=RESOLUTION_OPTIONS[0],
                label="Resolution",
            )

            seed = gr.Number(label="Seed (-1 for random)", value=-1, precision=0)
            generate_btn = gr.Button("Generate Image", variant="primary")
            seed_used = gr.Number(label="Seed Used", interactive=False)

        with gr.Column():
            output_image = gr.Image(label="Generated Image", type="pil")

    generate_btn.click(
        fn=generate_image,
        inputs=[prompt, resolution, seed],
        outputs=[output_image, seed_used],
    )

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