app.py

import gradio as gr
from diffusers import DiffusionPipeline
import dask
from dask import delayed, compute
from concurrent.futures import ThreadPoolExecutor
import os
os.environ['HF_HOME'] = '/blabla/cache/'


# Load model
pipe = DiffusionPipeline.from_pretrained("prompthero/openjourney-v4")

def generate_image(prompt, num_inference_steps=50):
    """
    Generate an image based on a text prompt using diffusion with optimizations.
    The number of inference steps is reduced for faster generation.
    """
    # Reduce steps for faster processing
    image = pipe(prompt, num_inference_steps=num_inference_steps).images[0]
    return image

@delayed
def dask_generate(prompt):
    return generate_image(prompt)

def parallel_generate(prompt):
    # Use multithreading to speed up the computation by processing multiple images simultaneously
    with ThreadPoolExecutor(max_workers=4) as executor:
        futures = [executor.submit(dask_generate, prompt) for _ in range(4)]  # Example with 4 threads
        results = [future.result() for future in futures]

    # Execute the generation using Dask to potentially improve processing speed
    images = compute(*results)
    return images[0]  # Return the first image generated for simplicity

# Gradio interface
iface = gr.Interface(
    fn=parallel_generate,
    inputs=gr.Textbox(label="Prompt", placeholder="Enter your prompt here"),
    outputs=gr.Image(type="pil"),
    title="Multithreaded CPU Optimized Image Generation",
    description="Enter a prompt to generate an image efficiently using CPU optimization and multithreading."
)

# Launch the Gradio app
if __name__ == "__main__":
    iface.launch()