app.py

import gradio as gr
import model_loader 
import pipeline
from transformers import CLIPTokenizer
import torch
import cv2
from download import download

DEVICE = "cpu"

ALLOW_CUDA = True
ALLOW_MPS = False

if torch.cuda.is_available() and ALLOW_CUDA:
    DEVICE = "cuda"
elif (torch.has_mps or torch.backends.mps.is_available()) and ALLOW_MPS:
    DEVICE = "mps"
print(f"Using device: {DEVICE}")

def generate_image(prompt):

    tokenizer = CLIPTokenizer(r"./data/vocab.json", merges_file="./data/merges.txt")
    model_file = "./data/v1-5-pruned-emaonly.ckpt"
    models = model_loader.preload_models_from_standard_weights(model_file, DEVICE)


    ## TEXT TO IMAGE
    uncond_prompt = ""  # Also known as negative prompt
    do_cfg = True
    cfg_scale = 8  # min: 1, max: 14

    ## SAMPLER

    sampler = "ddpm"
    num_inference_steps = 50
    seed = 42

    output_image = pipeline.generate(
        prompt=prompt,
        uncond_prompt=uncond_prompt,
        input_image=None,  
        strength=0.5,
        do_cfg=do_cfg,
        cfg_scale=cfg_scale,
        sampler_name=sampler,
        n_inference_steps=num_inference_steps,
        seed=seed,
        models=models,
        device=DEVICE,
        idle_device="cpu",  
        tokenizer=tokenizer,
    )
    
    image = cv2.cvtColor(output_image, cv2.COLOR_RGB2BGR)
    
    return image
    
    
with gr.Blocks() as demo:
    gr.Markdown('# Diffusion')
    with gr.Row():
        with gr.Column():
            # Add a text box
            prompt = gr.Textbox(lines=3, label="Prompt", placeholder="Enter a prompt")
            print(prompt)
        with gr.Column():
            output_image = gr.Image(label="Generated Image")

    submit_button_image = gr.Button("Generate Image")
    submit_button_image.click(generate_image, inputs=prompt, outputs=output_image)
    
if __name__ == "__main__":
    download()
    demo.launch()