import gradio as gr
import numpy as np
import random
import spaces
import torch
from diffusers import DiffusionPipeline

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

pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=dtype).to(device)

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

@spaces.GPU()
def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, num_inference_steps=4, progress=gr.Progress(track_tqdm=True)):
    lora = "kratadata/luisa-doodle"
    prefix = "A black and white pen doodle in the style of BOODLE of "
    suffix = "Completly white background. Doodle in the center, and it is not touching the edges."

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    generator = torch.Generator().manual_seed(seed)
    full_prompt = prefix + " " + prompt + " " + suffix
    pipe.load_lora_weights(lora)
    image = pipe(
                prompt = full_prompt,
                width = width,
                height = height,
                num_inference_steps = num_inference_steps,
                generator = generator,
                guidance_scale=0.0
    ).images[0]
    return image, seed


with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            prompt = gr.Textbox(
                label="Image Prompt",
                show_label = "True",
                info="Your image prompt",
                max_lines=4,
                placeholder="Enter your prompt",
                container=True,
            )
            with gr.Accordion("Advanced Settings", open=False):
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                    info = "Keep at 1024 for best results"
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                    info = "Keep at 1024 for best results"
                )
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=8,
                    step=1,
                    value=2,
                    info = "Increase to 4 for better results"
                )
                randomize_seed = gr.Checkbox(label="Randomize seed", value=True, info = "Keep true to generate a new image each time")
                seed = gr.Slider(
                    label="Seed",
                    minimum=0,
                    maximum=MAX_SEED,
                    step=1,
                    value=0,
                    info = "Fix seed if you want to keep generating the same image"
                )
            run_button = gr.Button("Run", scale=0)

        with gr.Column():
            result = gr.Image(label="Result", show_label=False, format="jpeg")



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

demo.launch()