app.py

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

# Ensure sentencepiece is installed in your environment
try:
    import sentencepiece
except ImportError:
    raise ImportError("The 'sentencepiece' library is required but not installed. Please add it to your environment.")

# Set the device and dtype
dtype = torch.float16  # Change to float16 for better compatibility and performance
device = "cuda" if torch.cuda.is_available() else "cpu"

# Load the diffusion pipeline without requiring an API token
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 = 2048

@spaces.GPU()
def infer(prompt, negative_prompt=None, seed=42, randomize_seed=False, width=1024, height=1024, num_inference_steps=4, guidance_scale=7.5, progress=gr.Progress(track_tqdm=True)):
    start_time = time.time()
    
    if width > MAX_IMAGE_SIZE or height > MAX_IMAGE_SIZE:
        raise ValueError("Image size exceeds the maximum allowed dimensions.")
    
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    generator = torch.Generator(device=device).manual_seed(seed)
    
    try:
        # Include negative prompts in the diffusion pipeline call
        image = pipe(
            prompt=prompt, 
            negative_prompt=negative_prompt,  # Using the negative prompt here
            width=width,
            height=height,
            num_inference_steps=num_inference_steps, 
            generator=generator,
            guidance_scale=guidance_scale
        ).images[0]
    except Exception as e:
        print(f"Error generating image: {e}")
        return None, seed, f"Error: {str(e)}"
    
    if time.time() - start_time > 60:  # 60 seconds timeout
        return None, seed, "Image generation took too long and was cancelled."

    return image, seed, None

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: 640px;
    padding: 20px;
    box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);
    border-radius: 10px;
    background-color: #f8f9fa;
}
#run-button {
    background-color: #007bff;
    color: white;
    border: none;
    padding: 10px 20px;
    font-size: 16px;
    border-radius: 5px;
}
#run-button:hover {
    background-color: #0056b3;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("""
        # Custom Image Creator
        A 12B param rectified flow transformer from [FLUX.1](https://blackforestlabs.ai/) for 4-step generation.
        """, elem_id="title")
        
        prompt = gr.Textbox(
            label="Prompt",
            show_label=False,
            max_lines=1,
            placeholder="Enter your prompt...",
        )
        negative_prompt = gr.Textbox(
            label="Negative Prompt",
            show_label=False,
            max_lines=1,
            placeholder="Enter negative prompts (what to avoid)...",
        )
        run_button = gr.Button("Run", elem_id="run-button")
        
        result = gr.Image(label="Result", show_label=False, interactive=True)
        
        with gr.Accordion("Advanced Settings", open=False):
            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
                tooltip="Seed value for reproducibility. Randomize for unique results."
            )
            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,
                    tooltip="Adjust the width of the generated image."
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                    tooltip="Adjust the height of the generated image."
                )
            
            with gr.Row():
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=4,
                    tooltip="Controls the quality and coherence of the output."
                )
                guidance_scale = gr.Slider(
                    label="Guidance Scale",
                    minimum=0.0,
                    maximum=20.0,
                    step=0.5,
                    value=7.5,
                    tooltip="Higher values result in outputs closer to the prompt."
                )
        
        gr.Examples(
            examples=examples,
            fn=infer,
            inputs=[prompt],
            outputs=[result, seed],
            cache_examples="lazy"
        )

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

    gr.Markdown("""
    ## Save Your Image
    Right-click on the image and select 'Save As' to download the generated image.
    """)
    
demo.launch()