app.py

import random
import re

import gradio as gr
import torch

from transformers import AutoModelForCausalLM
from transformers import AutoTokenizer
from transformers import AutoModelForSeq2SeqLM

from transformers import AutoProcessor

from transformers import pipeline

from transformers import set_seed

device = "cuda" if torch.cuda.is_available() else "cpu"

big_processor = AutoProcessor.from_pretrained("microsoft/git-base-coco")
big_model = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")

text_pipe = pipeline('text-generation', model='succinctly/text2image-prompt-generator')

zh2en_model = AutoModelForSeq2SeqLM.from_pretrained('Helsinki-NLP/opus-mt-zh-en').eval()
zh2en_tokenizer = AutoTokenizer.from_pretrained('Helsinki-NLP/opus-mt-zh-en')

en2zh_model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-zh").eval()
en2zh_tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-zh")


def translate_zh2en(text):
    with torch.no_grad():

        text = re.sub(r'([^\u4e00-\u9fa5])([\u4e00-\u9fa5])', r'\1\n\2', text)
        text = re.sub(r'([\u4e00-\u9fa5])([^\u4e00-\u9fa5])', r'\1\n\2', text)

        text = text.replace('\n', ',')

        text =re.sub(r'(?<![a-zA-Z])\s+|\s+(?![a-zA-Z])', '', text)

        text = re.sub(r',+', ',', text)

        encoded = zh2en_tokenizer([text], return_tensors='pt')
        sequences = zh2en_model.generate(**encoded)
        result = zh2en_tokenizer.batch_decode(sequences, skip_special_tokens=True)[0]

        result = result.strip() 

        return result


def translate_en2zh(text):
    with torch.no_grad():

        encoded = en2zh_tokenizer([text], return_tensors="pt")
        sequences = en2zh_model.generate(**encoded)
        return en2zh_tokenizer.batch_decode(sequences, skip_special_tokens=True)[0]

def test05(text):

    return text

def test06(text):

    return text


def text_generate(text):
    seed = random.randint(100, 1000000)
    set_seed(seed)

    text_in_english = translate_zh2en(text)
    result = ""
    for _ in range(6):
        sequences = text_pipe(text_in_english, max_length=random.randint(60, 90), num_return_sequences=8)
        list = []
        for sequence in sequences:


            line = sequence['generated_text'].strip()

            if line != text_in_english and len(line) > (len(text_in_english) + 4):

                list.append(translate_en2zh(line)+"\n")
                list.append(line+"\n")
                list.append("\n")

        result = "".join(list)

        result = re.sub('[^ ]+\.[^ ]+', '', result)

        result = result.replace('<', '').replace('>', '')

        if result != '':
            break
    return result


def load_prompter():
    prompter_model = AutoModelForCausalLM.from_pretrained("microsoft/Promptist")
    tokenizer = AutoTokenizer.from_pretrained("gpt2")
    tokenizer.pad_token = tokenizer.eos_token
    tokenizer.padding_side = "left"
    return prompter_model, tokenizer

prompter_model, prompter_tokenizer = load_prompter()

def generate_prompter(text):
    text = translate_zh2en(text)

    input_ids = prompter_tokenizer(text.strip()+" Rephrase:", return_tensors="pt").input_ids
    eos_id = prompter_tokenizer.eos_token_id
    outputs = prompter_model.generate(
        input_ids,
        do_sample=False,
        max_new_tokens=75,
        num_beams=3,
        num_return_sequences=3,
        eos_token_id=eos_id,
        pad_token_id=eos_id,
        length_penalty=-1.0
    )
    output_texts = prompter_tokenizer.batch_decode(outputs, skip_special_tokens=True)

    result = []
    for output_text in output_texts:

        output_text = output_text.replace('<', '').replace('>', '')
        output_text = output_text.split("Rephrase:", 1)[-1].strip()

        result.append(translate_en2zh(output_text)+"\n")
        result.append(output_text+"\n")
        result.append("\n")
    return "".join(result)

def combine_text(text):
    text01 = generate_prompter(text)
    text02 = text_generate(text)
    return text01,text02

def get_prompt_from_image(input_image):
    image = input_image.convert('RGB')
    pixel_values = big_processor(images=image, return_tensors="pt").to(device).pixel_values
    generated_ids = big_model.to(device).generate(pixel_values=pixel_values, max_length=50)
    generated_caption = big_processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
    result01 = generate_prompter(generated_caption)
    result02 = text_generate(generated_caption)
    return result01,result02

with gr.Blocks() as block:
    with gr.Column():
        with gr.Tab('工作區'):
            with gr.Row():
                input_text = gr.Textbox(lines=12, label='輸入文字', placeholder='在此输入文字...')
                input_image = gr.Image(type='pil')
            with gr.Row():
                txt_prompter_btn = gr.Button('文生文')
                pic_prompter_btn = gr.Button('圖生文')
            with gr.Row():
                Textbox_1 = gr.Textbox(lines=6, label='生成方式A')
            with gr.Row():
                Textbox_2 = gr.Textbox(lines=6, label='生成方式B')
        with gr.Tab('測試區'):
            with gr.Row():
                input_test01 = gr.Textbox(lines=2, label='中英翻譯', placeholder='在此输入文字...')
                test01_btn = gr.Button('執行')
                Textbox_test01 = gr.Textbox(lines=2, label='輸出結果')
            with gr.Row():
                input_test02 = gr.Textbox(lines=2, label='英中翻譯', placeholder='在此输入文字...')
                test02_btn = gr.Button('執行')
                Textbox_test02 = gr.Textbox(lines=2, label='輸出結果')
            with gr.Row():
                input_test03 = gr.Textbox(lines=2, label='SD模式', placeholder='在此输入文字...')
                test03_btn = gr.Button('執行')
                Textbox_test03 = gr.Textbox(lines=2, label='輸出結果')
            with gr.Row():
                input_test04 = gr.Textbox(lines=2, label='瞎掰模式', placeholder='在此输入文字...')
                test04_btn = gr.Button('執行')
                Textbox_test04 = gr.Textbox(lines=2, label='輸出結果')
            with gr.Row():
                input_test05 = gr.Textbox(lines=2, label='沒作用', placeholder='在此输入文字...')
                test05_btn = gr.Button('執行')
                Textbox_test05 = gr.Textbox(lines=2, label='輸出結果')
            with gr.Row():
                input_test06 = gr.Textbox(lines=2, label='沒作用', placeholder='在此输入文字...')
                test06_btn = gr.Button('執行')
                Textbox_test06 = gr.Textbox(lines=2, label='輸出結果')

    txt_prompter_btn.click(
        fn=combine_text,
        inputs=input_text, 
        outputs=[Textbox_1,Textbox_2]
    )

    pic_prompter_btn.click(
        fn=get_prompt_from_image,
        inputs=input_image,
        outputs=[Textbox_1,Textbox_2]
    )

    test01_btn.click(
        fn=translate_zh2en,
        inputs=input_test01,
        outputs=Textbox_test01
    )

    test02_btn.click(
        fn=translate_en2zh,
        inputs=input_test02,
        outputs=Textbox_test02
    )

    test03_btn.click(
        fn=generate_prompter,
        inputs=input_test03,
        outputs=Textbox_test03
    )

    test04_btn.click(
        fn=text_generate,
        inputs=input_test04,
        outputs=Textbox_test04
    )
    test05_btn.click(
        fn=test05,
        inputs=input_test05,
        outputs=Textbox_test05
    )
    test06_btn.click(
        fn=test06,
        inputs=input_test06,
        outputs=Textbox_test06
    )

block.queue(max_size=64).launch(show_api=False, enable_queue=True, debug=True, share=False, server_name='0.0.0.0')