import gradio as gr
import numpy as np
import random
import spaces
import torch

from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast

from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images

dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"

# Загружаем автоэнкодер и VAE
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
good_vae = AutoencoderKL.from_pretrained(
    "ifmain/UltraReal_Fine-Tune",
    subfolder="vae",
    torch_dtype=dtype
).to(device)

# Загружаем основной пайплайн
pipe = DiffusionPipeline.from_pretrained(
    "ifmain/UltraReal_Fine-Tune",
    torch_dtype=dtype,
    vae=taef1
).to(device)

torch.cuda.empty_cache()

# Подключаем LoRA
pipe.load_lora_weights("ifMain/realism")

pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 2048

@spaces.GPU(duration=75)
def infer(
    prompt,
    seed=42,
    randomize_seed=False,
    width=1280,
    height=732,
    guidance_scale=3.5,
    num_inference_steps=28,
    progress=gr.Progress(track_tqdm=True)
):
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

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

    for img in pipe.flux_pipe_call_that_returns_an_iterable_of_images(
            prompt=prompt,
            guidance_scale=guidance_scale,
            num_inference_steps=num_inference_steps,
            width=width,
            height=height,
            generator=generator,
            output_type="pil",
            good_vae=good_vae,
        ):
            yield img, seed

# Полные примеры с различными стилями и условиями съемки
full_examples = [
    ["d1g1cam, amateur photo, low-lit, Young woman, late 20s, casually dressed in an oversized pink T-shirt, outdoors, her gaze directed to the side, sad expression."],
    ["v8s, Dimly lit photo, grungy aesthetic, gritty urban, Los Angeles city on background, interior of muscle car driving at high speed, first-person perspective."],
    ["35mm film photo, high contrast, cinematic lighting, mid-20s man with messy dark hair and a leather jacket, standing under neon lights, rainy evening, water reflections on pavement."],
    ["Vintage Polaroid, warm and faded colors, soft focus. A child playing in a sunflower field, early morning sunlight filtering through the leaves, a dreamy nostalgic atmosphere."]
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 520px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(
            """# UltraReal Fine-Tune (Flux.1 Dev)  

            **🚀 Фотореализм нового уровня!**  
            Вышла 4-я версия **UltraReal Fine-Tune**, основанная на **Flux.1 Dev**.  
            Скачать можно тут: [Civitai](https://civitai.com/models/978314?modelVersionId=1413133)  

            **🚀 Next-level photorealism!**  
            The 4th version of **UltraReal Fine-Tune**, based on **Flux.1 Dev**, has been released.  
            You can download it here: [Civitai](https://civitai.com/models/978314?modelVersionId=1413133)  

            [[model](https://huggingface.co/black-forest-labs/FLUX.1-dev)] [[non-commercial license](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)] [[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)]
            """
        )

        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)

        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=732,
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1280,
                )

            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance Scale",
                    minimum=1,
                    maximum=15,
                    step=0.1,
                    value=3.5,
                )
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=28,
                )

        gr.Examples(
          examples=full_examples,
          fn=infer,
          inputs=[prompt],  # Теперь передаём только prompt
          outputs=[result],
          cache_examples=False
        )

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

demo.launch()