app.py

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

taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder="vae", torch_dtype=dtype).to(device)
pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=dtype, vae=taef1).to(device)
torch.cuda.empty_cache()

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

pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)

article_text = """
<div style="text-align: center;">
    <p>Enjoying the tool? Buy me a coffee and get exclusive prompt guides!</p>
    <p><i>Instantly unlock helpful tips for creating better prompts!</i></p>
    <div style="display: flex; justify-content: center;">
        <a href="https://piczify.lemonsqueezy.com/buy/0f5206fa-68e8-42f6-9ca8-4f80c587c83e">
            <img src="https://www.buymeacoffee.com/assets/img/custom_images/yellow_img.png" 
                 alt="Buy Me a Coffee" 
                 style="height: 40px; width: auto; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2); border-radius: 10px;">
        </a>
    </div>
</div>
"""

@spaces.GPU()
def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, guidance_scale=3.5, num_inference_steps=28, lora_id=None, lora_scale=0.95, 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

    # Handle LoRA loading
    # Load LoRA weights and prepare joint_attention_kwargs
    if lora_id:
        pipe.unload_lora_weights()
        pipe.load_lora_weights(lora_id)
        joint_attention_kwargs = {"scale": lora_scale}
    else:
        joint_attention_kwargs = None
    
    try:
        # Call the custom pipeline function with the correct keyword argument
        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,  # Assuming good_vae is defined elsewhere
            joint_attention_kwargs=joint_attention_kwargs,  # Fixed parameter name
        ):
            yield img, seed
    finally:
        # Unload LoRA weights if they were loaded
        if lora_id:
            pipe.unload_lora_weights()



examples = [
    "a tiny astronaut hatching from an egg on the moon",
    "a cat holding a sign that says hello world",
    "an anime illustration of a wiener schnitzel",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 960px;
}
.generate-btn {
    background: linear-gradient(90deg, #4B79A1 0%, #283E51 100%) !important;
    border: none !important;
    color: white !important;
}
.generate-btn:hover {
    transform: translateY(-2px);
    box-shadow: 0 5px 15px rgba(0,0,0,0.2);
}
"""

with gr.Blocks(css=css) as app:
    gr.HTML("<center><h1>FLUX.1-Dev with LoRA support</h1></center>")
    with gr.Column(elem_id="col-container"):
        with gr.Row():
            with gr.Column():
                with gr.Row():
                    text_prompt = gr.Textbox(label="Prompt", placeholder="Enter a prompt here", lines=3, elem_id="prompt-text-input")
                with gr.Row():
                    custom_lora = gr.Textbox(label="Custom LoRA", info="LoRA Hugging Face path (optional)", placeholder="multimodalart/vintage-ads-flux")
                with gr.Row():
                    with gr.Accordion("Advanced Settings", open=False):
                        with gr.Row():
                            lora_scale = gr.Slider(
                                label="LoRA Scale",
                                minimum=0,
                                maximum=2,
                                step=0.01,
                                value=0.95,
                            )
                            width = gr.Slider(label="Width", value=1024, minimum=64, maximum=1216, step=8)
                            height = gr.Slider(label="Height", value=1024, minimum=64, maximum=1216, step=8)
                        seed = gr.Slider(label="Seed", value=-1, minimum=-1, maximum=4294967296, step=1)
                        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
                        with gr.Row():
                            steps = gr.Slider(label="Inference steps steps", value=28, minimum=1, maximum=100, step=1)
                            cfg = gr.Slider(label="Guidance Scale", value=3.5, minimum=1, maximum=20, step=0.5)
                        # method = gr.Radio(label="Sampling method", value="DPM++ 2M Karras", choices=["DPM++ 2M Karras", "DPM++ SDE Karras", "Euler", "Euler a", "Heun", "DDIM"])

                with gr.Row():
                    # text_button = gr.Button("Run", variant='primary', elem_id="gen-button")
                    text_button = gr.Button("✨ Generate Image", variant='primary', elem_classes=["generate-btn"])
            with gr.Column():
                with gr.Row():
                    image_output = gr.Image(type="pil", label="Image Output", elem_id="gallery")
                with gr.Row():
                    seed_output = gr.Textbox(label="Seed Used", show_copy_button = True)
        
        gr.Markdown(article_text)
        with gr.Column():
            gr.Examples(
                examples = examples,
                inputs = [text_prompt],
            )

        
        # text_button.click(query, inputs=[custom_lora, text_prompt, steps, cfg, randomize_seed, seed, width, height], outputs=[image_output,seed_output, seed])
        text_button.click(infer, inputs=[text_prompt, seed, randomize_seed, width, height, cfg, steps, custom_lora, lora_scale], outputs=[image_output,seed_output, seed])

app.launch(share=True)