app.py

import os
import pandas as pd
import torch
import gc
import re
import random
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer
from diffusers import StableDiffusionPipeline
import gradio as gr

# Initialize the text generation pipeline with the pre-quantized 8-bit model
model_name = 'HuggingFaceTB/SmolLM-1.7B-Instruct'
model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
text_generator = pipeline('text-generation', model=model, tokenizer=tokenizer, device=-1)  # Use CPU

# Load the Stable Diffusion model
model_id = "stabilityai/stable-diffusion-2-1-base"  # Smaller model
pipe = StableDiffusionPipeline.from_pretrained(model_id)
pipe = pipe.to("cpu")  # Use CPU

# Create a directory to save the generated images
output_dir = 'generated_images'
os.makedirs(output_dir, exist_ok=True)
os.chmod(output_dir, 0o777)

# Function to generate a detailed visual description prompt
def generate_description_prompt(user_prompt, user_examples):
    prompt = f'generate enclosed in quotes in the format "<description>" description according to guidelines of {user_prompt} different from {user_examples}'
    try:
        generated_text = text_generator(prompt, max_length=150, num_return_sequences=1, truncation=True)[0]['generated_text']
        match = re.search(r'"(.*?)"', generated_text)
        if match:
            generated_description = match.group(1).strip()  # Capture the description between quotes
            return f'"{generated_description}"'
        else:
            return None
    except Exception as e:
        print(f"Error generating description for prompt '{user_prompt}': {e}")
        return None

# Seed words pool
seed_words = []

used_words = set()

def generate_description(user_prompt, user_examples_list):
    seed_words.extend(user_examples_list)

    # Select a subject that has not been used
    available_subjects = [word for word in seed_words if word not in used_words]
    if not available_subjects:
        print("No more available subjects to use.")
        return None, None

    subject = random.choice(available_subjects)
    generated_description = generate_description_prompt(user_prompt, subject)
    
    if generated_description:
        # Remove any offending symbols
        clean_description = generated_description.encode('ascii', 'ignore').decode('ascii')

        # Print the generated description to the command line
        print(f"Generated description for subject '{subject}': {clean_description}")

        # Update used words and seed words
        used_words.add(subject)
        seed_words.append(clean_description.strip('"'))  # Add the generated description to the seed bank array without quotes

        return clean_description, subject
    else:
        return None, None

# Function to generate an image based on the description
def generate_image(description, seed=42):
    prompt = f'detailed photorealistic full shot of {description}'
    generator = torch.Generator().manual_seed(seed)
    image = pipe(
        prompt=prompt,
        width=512,
        height=512,
        num_inference_steps=10,  # Use 10 inference steps
        generator=generator,
        guidance_scale=7.5,
    ).images[0]
    return image

# Gradio interface
def gradio_interface(user_prompt, user_examples):
    user_examples_list = [example.strip().strip('"') for example in user_examples.split(',')]
    generated_description, subject = generate_description(user_prompt, user_examples_list)
    
    if generated_description:
        # Generate image
        image = generate_image(generated_description)
        image_path = os.path.join(output_dir, f"image_{len(os.listdir(output_dir))}.png")
        image.save(image_path)
        os.chmod(image_path, 0o777)
        
        return image, generated_description
    else:
        return None, "Failed to generate description."

iface = gr.Interface(
    fn=gradio_interface,
    inputs=[
        gr.Textbox(lines=2, placeholder="Enter the generation task or general thing you are looking for"),
        gr.Textbox(lines=2, placeholder='Provide a few examples (enclosed in quotes and separated by commas)')
    ],
    outputs=[
        gr.Image(label="Generated Image"),
        gr.Textbox(label="Generated Description")
    ],
    title="Description and Image Generator",
    description="Generate detailed descriptions and images based on your input."
)

iface.launch(server_name="0.0.0.0", server_port=7860)

# Clear GPU memory when the process is closed
def clear_gpu_memory():
    torch.cuda.empty_cache()
    gc.collect()
    print("GPU memory cleared.")