Spaces:
Paused
Paused
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
|
| 3 |
+
import torch
|
| 4 |
+
|
| 5 |
+
# تحميل النموذج والتوكنايزر
|
| 6 |
+
model_id = "methodya/arabic-summarizer-philosophy-v2"
|
| 7 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
|
| 8 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 9 |
+
|
| 10 |
+
# نقل النموذج إلى GPU إذا كان متوفراً
|
| 11 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
| 12 |
+
model = model.to(device)
|
| 13 |
+
model.eval()
|
| 14 |
+
|
| 15 |
+
def summarize(text, max_length=200, num_beams=7, length_penalty=1.2):
|
| 16 |
+
# تحويل القيم إلى النوع المناسب
|
| 17 |
+
max_length = int(max_length)
|
| 18 |
+
num_beams = int(num_beams)
|
| 19 |
+
length_penalty = float(length_penalty)
|
| 20 |
+
|
| 21 |
+
if not text.strip():
|
| 22 |
+
return "الرجاء إدخال نص للتلخيص"
|
| 23 |
+
|
| 24 |
+
# تجهيز البيانات للنموذج
|
| 25 |
+
inputs = tokenizer(
|
| 26 |
+
text,
|
| 27 |
+
return_tensors="pt",
|
| 28 |
+
max_length=2048,
|
| 29 |
+
padding=True,
|
| 30 |
+
truncation=True
|
| 31 |
+
)
|
| 32 |
+
inputs = {k: v.to(device) for k, v in inputs.items()}
|
| 33 |
+
|
| 34 |
+
# توليد الملخص مع معاملات محسنة
|
| 35 |
+
with torch.no_grad():
|
| 36 |
+
outputs = model.generate(
|
| 37 |
+
**inputs,
|
| 38 |
+
max_length=max_length,
|
| 39 |
+
min_length=30,
|
| 40 |
+
num_beams=num_beams,
|
| 41 |
+
length_penalty=length_penalty,
|
| 42 |
+
repetition_penalty=2.5,
|
| 43 |
+
do_sample=False,
|
| 44 |
+
temperature=1.0,
|
| 45 |
+
top_p=0.95,
|
| 46 |
+
no_repeat_ngram_size=4,
|
| 47 |
+
num_return_sequences=1,
|
| 48 |
+
early_stopping=True
|
| 49 |
+
)
|
| 50 |
+
|
| 51 |
+
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 52 |
+
|
| 53 |
+
# تعريف واجهة Gradio
|
| 54 |
+
interface = gr.Interface(
|
| 55 |
+
fn=summarize,
|
| 56 |
+
inputs=[
|
| 57 |
+
gr.Textbox(lines=8, label="النص", placeholder="أدخل النص العربي هنا..."),
|
| 58 |
+
gr.Slider(100, 400, value=200, label="طول الملخص"),
|
| 59 |
+
gr.Slider(1, 10, value=7, step=1, label="دقة التلخيص (num_beams)"),
|
| 60 |
+
gr.Slider(0.5, 2.0, value=1.2, step=0.1, label="معامل الطول (length_penalty)")
|
| 61 |
+
],
|
| 62 |
+
outputs=gr.Textbox(label="الملخص"),
|
| 63 |
+
title="نظام تلخيص النصوص الفلسفية العربية",
|
| 64 |
+
description="نموذج متقدم لتلخيص النصوص الفلسفية باللغة العربية",
|
| 65 |
+
submit_btn="تلخيص",
|
| 66 |
+
clear_btn="مسح"
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
# تشغيل التطبيق
|
| 70 |
+
interface.launch()
|