Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer
|
3 |
+
import torch
|
4 |
+
from langdetect import detect, LangDetectException
|
5 |
+
|
6 |
+
# Load model dan tokenizer
|
7 |
+
model_name = "facebook/m2m100_418M"
|
8 |
+
try:
|
9 |
+
tokenizer = M2M100Tokenizer.from_pretrained(model_name)
|
10 |
+
model = M2M100ForConditionalGeneration.from_pretrained(model_name)
|
11 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
12 |
+
model.to(device)
|
13 |
+
model.eval()
|
14 |
+
except Exception as e:
|
15 |
+
raise Exception(f"Gagal memuat model: {str(e)}")
|
16 |
+
|
17 |
+
# Fungsi terjemahan
|
18 |
+
def translate_text(text, source_lang=None):
|
19 |
+
try:
|
20 |
+
# Autodeteksi bahasa jika source_lang tidak diberikan
|
21 |
+
if not source_lang:
|
22 |
+
try:
|
23 |
+
detected_lang = detect(text)
|
24 |
+
if detected_lang not in tokenizer.supported_languages:
|
25 |
+
return f"Bahasa terdeteksi '{detected_lang}' tidak didukung.", detected_lang
|
26 |
+
source_lang = detected_lang
|
27 |
+
except LangDetectException:
|
28 |
+
return "Gagal mendeteksi bahasa. Harap masukkan kode bahasa sumber.", None
|
29 |
+
else:
|
30 |
+
if source_lang not in tokenizer.supported_languages:
|
31 |
+
return f"Kode bahasa '{source_lang}' tidak didukung.", None
|
32 |
+
|
33 |
+
# Set bahasa sumber
|
34 |
+
tokenizer.src_lang = source_lang
|
35 |
+
|
36 |
+
# Encode dan terjemahkan
|
37 |
+
encoded = tokenizer(text, return_tensors="pt", padding=True, truncation=True).to(device)
|
38 |
+
generated_tokens = model.generate(
|
39 |
+
**encoded,
|
40 |
+
forced_bos_token_id=tokenizer.get_lang_id("en")
|
41 |
+
)
|
42 |
+
translated_text = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
|
43 |
+
|
44 |
+
return translated_text, source_lang
|
45 |
+
|
46 |
+
except Exception as e:
|
47 |
+
return f"Error: {str(e)}", None
|
48 |
+
|
49 |
+
# Buat antarmuka Gradio
|
50 |
+
iface = gr.Interface(
|
51 |
+
fn=translate_text,
|
52 |
+
inputs=[
|
53 |
+
gr.Textbox(label="Teks untuk Diterjemahkan"),
|
54 |
+
gr.Dropdown(
|
55 |
+
choices=["id", "fr", "es", "de", "ja", ""], # Tambahkan lebih banyak kode bahasa jika perlu
|
56 |
+
label="Bahasa Sumber (kosongkan untuk autodeteksi)",
|
57 |
+
value=""
|
58 |
+
)
|
59 |
+
],
|
60 |
+
outputs=[
|
61 |
+
gr.Textbox(label="Terjemahan ke Bahasa Inggris"),
|
62 |
+
gr.Textbox(label="Bahasa Sumber Terdeteksi")
|
63 |
+
],
|
64 |
+
title="M2M100 Translation to English",
|
65 |
+
description="Masukkan teks untuk diterjemahkan ke bahasa Inggris. Biarkan bahasa sumber kosong untuk autodeteksi."
|
66 |
+
)
|
67 |
+
|
68 |
+
# Luncurkan aplikasi
|
69 |
+
if __name__ == "__main__":
|
70 |
+
iface.launch()
|