Spaces:
Paused
Paused
Update src/main.py
Browse files- src/main.py +4 -9
src/main.py
CHANGED
@@ -2,7 +2,7 @@ import display_gloss as dg
|
|
2 |
import synonyms_preprocess as sp
|
3 |
from NLP_Spacy_base_translator import NlpSpacyBaseTranslator
|
4 |
from flask import Flask, render_template, Response, request
|
5 |
-
from transformers import
|
6 |
import torch
|
7 |
import os
|
8 |
|
@@ -22,8 +22,8 @@ os.environ['CUDA_VISIBLE_DEVICES'] = ''
|
|
22 |
|
23 |
# Load pre-trained Korean-English translation model
|
24 |
model_name = "Helsinki-NLP/opus-mt-ko-en"
|
25 |
-
tokenizer =
|
26 |
-
model =
|
27 |
model = model.to(device)
|
28 |
|
29 |
nlp, dict_docs_spacy = sp.load_spacy_values()
|
@@ -31,12 +31,11 @@ dataset, list_2000_tokens = dg.load_data()
|
|
31 |
|
32 |
def translate_korean_to_english(text):
|
33 |
try:
|
34 |
-
# Check if input is Korean
|
35 |
if any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in text):
|
36 |
inputs = tokenizer(text, return_tensors="pt", padding=True)
|
37 |
outputs = model.generate(**inputs)
|
38 |
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
39 |
-
print(f"Translated text: {translation}")
|
40 |
return translation
|
41 |
return text
|
42 |
except Exception as e:
|
@@ -52,18 +51,14 @@ def result():
|
|
52 |
if request.method == 'POST':
|
53 |
input_text = request.form['inputSentence']
|
54 |
try:
|
55 |
-
# Translate to English
|
56 |
english_text = translate_korean_to_english(input_text)
|
57 |
|
58 |
-
# Check if translation failed
|
59 |
if english_text == input_text and any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in input_text):
|
60 |
raise Exception("Translation failed")
|
61 |
|
62 |
-
# Convert to ASL gloss
|
63 |
eng_to_asl_translator = NlpSpacyBaseTranslator(sentence=english_text)
|
64 |
generated_gloss = eng_to_asl_translator.translate_to_gloss()
|
65 |
|
66 |
-
# Process gloss
|
67 |
gloss_list_lower = [gloss.lower() for gloss in generated_gloss.split() if gloss.isalnum()]
|
68 |
gloss_sentence_before_synonym = " ".join(gloss_list_lower)
|
69 |
|
|
|
2 |
import synonyms_preprocess as sp
|
3 |
from NLP_Spacy_base_translator import NlpSpacyBaseTranslator
|
4 |
from flask import Flask, render_template, Response, request
|
5 |
+
from transformers import MarianMTModel, MarianTokenizer
|
6 |
import torch
|
7 |
import os
|
8 |
|
|
|
22 |
|
23 |
# Load pre-trained Korean-English translation model
|
24 |
model_name = "Helsinki-NLP/opus-mt-ko-en"
|
25 |
+
tokenizer = MarianTokenizer.from_pretrained(model_name, cache_dir=cache_dir)
|
26 |
+
model = MarianMTModel.from_pretrained(model_name, cache_dir=cache_dir)
|
27 |
model = model.to(device)
|
28 |
|
29 |
nlp, dict_docs_spacy = sp.load_spacy_values()
|
|
|
31 |
|
32 |
def translate_korean_to_english(text):
|
33 |
try:
|
|
|
34 |
if any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in text):
|
35 |
inputs = tokenizer(text, return_tensors="pt", padding=True)
|
36 |
outputs = model.generate(**inputs)
|
37 |
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
38 |
+
print(f"Translated text: {translation}")
|
39 |
return translation
|
40 |
return text
|
41 |
except Exception as e:
|
|
|
51 |
if request.method == 'POST':
|
52 |
input_text = request.form['inputSentence']
|
53 |
try:
|
|
|
54 |
english_text = translate_korean_to_english(input_text)
|
55 |
|
|
|
56 |
if english_text == input_text and any('\u3131' <= char <= '\u318F' or '\uAC00' <= char <= '\uD7A3' for char in input_text):
|
57 |
raise Exception("Translation failed")
|
58 |
|
|
|
59 |
eng_to_asl_translator = NlpSpacyBaseTranslator(sentence=english_text)
|
60 |
generated_gloss = eng_to_asl_translator.translate_to_gloss()
|
61 |
|
|
|
62 |
gloss_list_lower = [gloss.lower() for gloss in generated_gloss.split() if gloss.isalnum()]
|
63 |
gloss_sentence_before_synonym = " ".join(gloss_list_lower)
|
64 |
|