File size: 2,298 Bytes
4fd8256
 
 
 
 
 
2a9d3fa
 
 
4fd8256
cffbd0f
 
 
 
 
4fd8256
cffbd0f
 
4fd8256
 
 
 
 
2ab56af
2a9d3fa
 
 
2854501
d19bce7
2a9d3fa
ec0b50d
db6849b
ec0b50d
 
cffbd0f
 
2a9d3fa
cffbd0f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ec0b50d
cffbd0f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4fd8256
cffbd0f
 
 
 
 
 
4fd8256
ec0b50d
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
# -*- coding: utf-8 -*-
import numpy as np
import streamlit as st
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM


st.set_page_config(
    page_title="KoQuillBot", layout="wide", initial_sidebar_state="expanded"
)

@st.cache
def load_model(model_name):
    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
    return model

tokenizer = AutoTokenizer.from_pretrained("QuoQA-NLP/KE-T5-Ko2En-Base")
ko2en_model = load_model("QuoQA-NLP/KE-T5-Ko2En-Base")
en2ko_model = load_model("QuoQA-NLP/KE-T5-En2Ko-Base")


st.title("πŸ€– KoQuillBot")


default_value = "ν”„λ‘œμ νŠΈ κ°€μΉ˜κ°€ λ―Έν™” 1백만 λ‹¬λŸ¬ 이상인 곡곡 νŒŒνŠΈλ„ˆκ°€ μ‹œμž‘ν•œ PPP ν”„λ‘œμ νŠΈμ— λŒ€ν•΄ 2단계 μž…μ°°μ΄ μ‹€μ‹œλ©λ‹ˆλ‹€. μž…μ°°μ„ μ „μž λ°©μ‹μœΌλ‘œ μ§„ν–‰ν•˜λŠ” 것이 ν—ˆμš©λ©λ‹ˆλ‹€. (즉, μ‹ μ²­μ„œ 및 μž…μ°° μ œμ•ˆμ˜ μ „μž 제좜). COVID-19 전염병과 그에 λ”°λ₯Έ μ—¬ν–‰ μ œν•œμœΌλ‘œ 인해 μ˜€λŠ˜λ‚ μ—λŠ” 일반적인 관행이 λ˜μ—ˆμŠ΅λ‹ˆλ‹€."
src_text = st.text_area(
    "λ°”κΎΈκ³  싢은 λ¬Έμž₯을 μž…λ ₯ν•˜μ„Έμš”:",
    default_value,
    height=300,
    max_chars=200,
)
print(src_text)



if src_text == "":
    st.warning("Please **enter text** for translation")

# translate into english sentence
english_translation = ko2en_model.generate(
    **tokenizer(
        src_text,
        return_tensors="pt",
        padding="max_length",
        truncation=True,
        max_length=64,
    ),
    max_length=64,
    num_beams=5,
    repetition_penalty=1.3,
    no_repeat_ngram_size=3,
    num_return_sequences=1,
)
english_translation = tokenizer.decode(
    english_translation[0],
    clean_up_tokenization_spaces=True,
    skip_special_tokens=True,
)

# translate back to korean
korean_translation = en2ko_model.generate(
    **tokenizer(
        english_translation,
        return_tensors="pt",
        padding="max_length",
        truncation=True,
        max_length=64,
    ),
    max_length=64,
    num_beams=5,
    repetition_penalty=1.3,
    no_repeat_ngram_size=3,
    num_return_sequences=1,
)

korean_translation = tokenizer.decode(
    korean_translation[0],
    clean_up_tokenization_spaces=True,
    skip_special_tokens=True,
)
print(f"{src_text} -> {english_translation} -> {korean_translation}")

st.write(korean_translation)
print(korean_translation)