app.py

from huggingface_hub import InferenceClient
from PIL import Image,ImageFont,ImageDraw
import gradio as gr
import requests
import random
import uuid
import io
from utils import models, MEME_GENERATOR,GENERATE_PROMPT
client = InferenceClient(
    "mistralai/Mixtral-8x7B-Instruct-v0.1"
)
loaded_model=[]
for i,model in enumerate(models):
    loaded_model.append(gr.load(f'models/{model}', cache_examples=False))

def textover(im,text):
    x=0
    y=0
    t_fill = (0,0,0)
    font_size=20
    
    draw = ImageDraw.Draw(im)
    
    font = ImageFont.truetype("./fonts/unifont-15.0.01.ttf", int(font_size))
    draw.text((x, y),text, font = font, fill=t_fill)
    return im
def get_concat_h_cut(in1, in2):
    im1=Image.open(in1)
    im2=Image.open(in2)
    dst = Image.new('RGB', (im1.width + im2.width,
                            min(im1.height, im2.height)))
    dst.paste(im1, (0, 0))
    dst.paste(im2, (im1.width, 0))
    return dst


def get_concat_v_cut(im1,theme='light'):
    #im1=Image.open(in1)
    if theme=='dark':
        color=(31,41,55)
    if theme=='light':
        color=(255,255,255)
    dst = Image.new('RGB', (im1.width, im1.height +200),color=color)
    dst.paste(im1, (0, 200))
    return dst






def format_prompt(message, history):
  prompt = "<s>"
  for user_prompt, bot_response in history:
    prompt += f"[INST] {user_prompt} [/INST]"
    prompt += f" {bot_response}</s> "
  prompt += f"[INST] {message} [/INST]"
  return prompt
agents =[
    "MEME_GENERATOR",
]
def generate(prompt, history):
    print(f'HISTORY:: {history}')
    history=[]
    output1={}
    seed = random.randint(1,1111111111111111)

    system_prompt=MEME_GENERATOR
  
    generate_kwargs = dict(
        temperature=0.7,
        max_new_tokens=256,
        top_p=0.95,
        repetition_penalty=1,
        do_sample=True,
        seed=seed,
    )

    formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
        yield [(prompt,output)],output1
  
    if "PROMPT:" and "MEME_TEXT:" in output:
        print("YES")
        prompt_t=output.split("PROMPT:",1)[1].split("MEME_TEXT:",1)[0].strip()
        print(prompt_t)
        meme_t=output.split("MEME_TEXT:",1)[1].strip()
        print(meme_t)
        output1={'PROMPT':prompt_t,'MEME_TEXT':meme_t}
        #output=str(output1)
    yield [(prompt,output)],output1
    
def run(inp,model_drop):
    prompt=inp['PROMPT']
    text=inp['MEME_TEXT']
    model=loaded_model[int(model_drop)]
    out_img=model(prompt)
    print(out_img)
    url=f'https://omnibus-meme-diffusion.hf.space/file={out_img}'
    print(url)
    uid = uuid.uuid4()
    r = requests.get(url, stream=True)
    if r.status_code == 200:
        out = Image.open(io.BytesIO(r.content))
    out=get_concat_v_cut(out)
    out=textover(out,text)
    return out

def run_gpt(in_prompt,history,):
    if len(in_prompt)>max_prompt:
        in_prompt = condense(in_prompt)
    print(f'history :: {history}')
    prompt=format_prompt(in_prompt,history)
    seed = random.randint(1,1111111111111111)
    print (seed)
    generate_kwargs = dict(
        temperature=1.0,
        max_new_tokens=1048,
        top_p=0.99,
        repetition_penalty=1.0,
        do_sample=True,
        seed=seed,
    )
    content = GENERATE_PROMPT + prompt
    print(content)
    stream = client.text_generation(content, **generate_kwargs, stream=True, details=True, return_full_text=False)
    resp = ""
    for response in stream:
        resp += response.token.text
    return resp


    
with gr.Blocks() as app:
    gr.HTML("""<center><h1>Meme Diffusion</h1></center>""")
    with gr.Row():
        with gr.Column(scale=1):
            chatbot=gr.Chatbot()
            msg = gr.Textbox()
            model_drop=gr.Dropdown(label="Diffusion Models", type="index", choices=[m for m in models], value=models[0])
            with gr.Group():
                submit_b = gr.Button("Meme")
                submit_im = gr.Button("Image")
                with gr.Row():
                    stop_b = gr.Button("Stop")
                    clear = gr.ClearButton([msg, chatbot])
        with gr.Column(scale=2):
            im_out=gr.Image(label="Image")
    json_out=gr.JSON()
        
    sub_b = submit_b.click(generate, [msg,chatbot],[chatbot,json_out])
    sub_im = submit_im.click(run, [json_out,model_drop],[im_out])
    stop_b.click(None,None,None, cancels=[sub_b,sub_im])
app.launch()