app.py

import os
import re
import spaces
import random
import string
import torch
import requests
import gradio as gr
import numpy as np
from lxml.html import fromstring
from transformers import pipeline
from torch.multiprocessing import Pool, Process, set_start_method
#from pathos.multiprocessing import ProcessPool as Pool
#from pathos.threading import ThreadPool as Pool
#from diffusers.pipelines.flux import FluxPipeline
#from diffusers.utils import export_to_gif
#from huggingface_hub import hf_hub_download
#from safetensors.torch import load_file
from diffusers import DiffusionPipeline, StableDiffusionXLImg2ImgPipeline
from diffusers.utils import load_image
#import jax
#import jax.numpy as jnp
import torch._dynamo

set_start_method("spawn", force=True)
torch._dynamo.config.suppress_errors = True

#pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16, revision="refs/pr/1", token=os.getenv("hf_token")).to(device)
#pipe2 = StableDiffusionXLImg2ImgPipeline.from_pretrained("stabilityai/stable-diffusion-xl-refiner-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True).to(device)
#pipe2.unet = torch.compile(pipe2.unet, mode="reduce-overhead", fullgraph=True)

PIPE = None

def pipe_t2i():
    global PIPE
    if PIPE is None:
        PIPE = pipeline("text-to-image", model="black-forest-labs/FLUX.1-schnell", torch_dtype=torch.bfloat16, revision="refs/pr/1", tokenizer="black-forest-labs/FLUX.1-schnell", device=-1, token=os.getenv("hf_token"))
    return PIPE
    
def pipe_i2i():
    global PIPE
    if PIPE is None:
        PIPE = pipeline("image-to-image", model="stabilityai/stable-diffusion-xl-refiner-1.0", tokenizer="stabilityai/stable-diffusion-xl-refiner-1.0", torch_dtype=torch.float16, device=-1, variant="fp16", use_safetensors=True)
        PIPE.unet = torch.compile(PIPE.unet, mode="reduce-overhead", fullgraph=True)
    return PIPE
    
def translate(text,lang):
    if text == None or lang == None:
        return ""       
    text = re.sub(f'[{string.punctuation}]', '', re.sub('[\s+]', ' ', text)).lower().strip()
    lang = re.sub(f'[{string.punctuation}]', '', re.sub('[\s+]', ' ', lang)).lower().strip()    
    if text == "" or lang == "":
        return ""
    if len(text) > 38:
        raise Exception("Translation Error: Too long text!")
    user_agents = [
        'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36',
        'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36',
        'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36',
        'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.1 Safari/605.1.15',
        'Mozilla/5.0 (Macintosh; Intel Mac OS X 13_1) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.1 Safari/605.1.15'
    ]
    padded_chars = re.sub("[(^\-)(\-$)]","",text.replace("","-").replace("- -"," ")).strip()
    query_text = f'Please translate {padded_chars}, into {lang}'
    url = f'https://www.google.com/search?q={query_text}'
    resp = requests.get(
        url = url,
        headers = {
            'User-Agent': random.choice(user_agents)
        }
    )
    content = resp.content
    html = fromstring(content)
    translated = text
    try:
        src_lang = html.xpath('//*[@class="source-language"]')[0].text_content().lower().strip()
        trgt_lang = html.xpath('//*[@class="target-language"]')[0].text_content().lower().strip()
        src_text = html.xpath('//*[@id="tw-source-text"]/*')[0].text_content().lower().strip()
        trgt_text = html.xpath('//*[@id="tw-target-text"]/*')[0].text_content().lower().strip()
        if trgt_lang == lang:
            translated = trgt_text
    except:
        print(f'Translation Warning: Failed To Translate!')
    ret = re.sub(f'[{string.punctuation}]', '', re.sub('[\s+]', ' ', translated)).lower().strip()
    print(ret)
    return ret

def generate_random_string(length):
    characters = string.ascii_letters + string.digits
    return ''.join(random.choice(characters) for _ in range(length))

@spaces.GPU(duration=35)
def Piper(_do):
    pipe = pipe_t2i()
    try:
        retu = pipe(
            _do,
            height=512,
            width=512,
            num_inference_steps=4,
            max_sequence_length=256,
            guidance_scale=0
        )
        return retu
    except Exception as e:
        print(e)
        return None

@spaces.GPU(duration=35)
def Piper2(img,posi,neg):
    pipe = pipe_i2i()
    try:
        retu = pipe2(
            prompt=posi,
            negative_prompt=neg,
            image=img
        )
        return retu
    except Exception as e:
        print(e)
        return None

@spaces.GPU(duration=35)
def tok(txt):
    toks = pipe.tokenizer(txt)['input_ids']
    print(toks)
    return toks

def infer(p1,p2):
    name = generate_random_string(12)+".png"
    _do = ['beautiful', 'playful', 'photographed', 'realistic', 'dynamic poze', 'deep field', 'reasonable coloring', 'rough texture', 'best quality', 'focused']
    if p1 != "":
        _do.append(f'{p1}')
    if p2 != "":
        _dont = f'{p2} where in {p1}'
        neg = _dont
    else:
        neg = None
    output = Piper('A '+" ".join(_do))
    if output == None:
        return None
    else:
        output.images[0].save(name)
    if neg == None:
        return name

    img = load_image(name).convert("RGB")
    output2 = Piper2(img,p1,neg)
    if output2 == None:
        return None
    else:
        output2.images[0].save("_"+name)
    return "_"+name

css="""
input, input::placeholder {
    text-align: center !important;
}
*, *::placeholder {
    direction: ltr !important;
    font-family: Suez One !important;
}
h1,h2,h3,h4,h5,h6,span,p,pre {
    width: 100% !important;
    text-align: center !important;
    display: block !important;
}
footer {
    display: none !important;
}
#col-container {
    margin: 0 auto !important;
    max-width: 15cm !important;
 }
.image-container {
    aspect-ratio: 448 / 448 !important;
}
.dropdown-arrow {
    display: none !important;
}
*:has(.btn), .btn {
    width: 100% !important;
    margin: 0 auto !important;
}
"""

js="""
function custom(){
    document.querySelector("div#prompt input").setAttribute("maxlength","38")
    document.querySelector("div#prompt2 input").setAttribute("maxlength","38")
}
"""

with gr.Blocks(theme=gr.themes.Soft(),css=css,js=js) as demo:
    result = []
    with gr.Column(elem_id="col-container"):
        gr.Markdown(f"""
            # MULTI-LANGUAGE IMAGE GENERATOR
        """)
        with gr.Row():
            prompt = gr.Textbox(
                elem_id="prompt",
                placeholder="INCLUDE",
                container=False,
                max_lines=1
            )
        with gr.Row():
            prompt2 = gr.Textbox(
                elem_id="prompt2",
                placeholder="EXCLUDE",
                container=False,
                max_lines=1
            )
        with gr.Row():
            run_button = gr.Button("START",elem_classes="btn",scale=0)
        with gr.Row():
            result.append(gr.Image(interactive=False,elem_classes="image-container", label="Result", show_label=False, type='filepath', show_share_button=False))
            result.append(gr.Image(interactive=False,elem_classes="image-container", label="Result", show_label=False, type='filepath', show_share_button=False))
            result.append(gr.Image(interactive=False,elem_classes="image-container", label="Result", show_label=False, type='filepath', show_share_button=False))

    def _ret(p):

        print(f'Starting!')
        v = infer(p["a"],p["b"])
        print(f'Finished!')
        return v
        
    def _rets(p1,p2):
        
        p1_en = translate(p1,"english")
        p2_en = translate(p2,"english")
        
        p = {"a":p1_en,"b":p2_en}
        
        ln = len(result)
        rng = range(ln)
        p_arr = [p for _ in rng]
        pool = Pool(processes=ln)
        lst = list( pool.imap( _ret, p_arr ) )
        pool.clear()
        return lst
        
        #return list( _ret(p1_en,p2_en) )
        
    run_button.click(fn=_rets,inputs=[prompt,prompt2],outputs=result)

demo.queue().launch(server_port=6900)