Update app.py

app.py

@@ -1,54 +1,23 @@
 import tensorflow as tf
-import numpy as np
-import gradio as gr
 from transformers import BertTokenizer
 
-# تحميل الـ Tokenizer
+# تحميل tokenizer من الملفات اللي في نفس المسار
 tokenizer = BertTokenizer.from_pretrained(".")
 
 # تحميل الموديل
 model = tf.keras.models.load_model("rnn_Bi.h5")
-
-# دالة التنبؤ
-def predict_sentiment(text):
-    if not text.strip():
-        return "⛔ Please enter some text."
-
-    # تجهيز الإدخال
-    tokens = tokenizer(
-        text,
-        padding='max_length',
-        truncation=True,
-        max_length=128,
-        return_tensors="np"
-    )
-    input_ids = tokens["input_ids"]
-    attention_mask = tokens["attention_mask"]
-
-    # التنبؤ
-    prediction = model.predict([input_ids, attention_mask])[0][0]
-    label = int(prediction > 0.5)
-
-    # تحويل النتيجة لنص
-    if label == 1:
-        return "🌟 Positive Sentiment 😊"
-    else:
-        return "😞 Negative Sentiment"
-
-# واجهة Gradio
-interface = gr.Interface(
-    fn=predict_sentiment,
-    inputs=gr.Textbox(lines=3, placeholder="Write your sentence here...", label="Enter Text"),
-    outputs=gr.Textbox(label="Prediction"),
-    title="Sentiment Analysis - RNN BiLSTM",
-    description="This model predicts whether the input sentence has a positive or negative sentiment.",
-    theme="soft",
-    examples=[
-        ["I love this product!"],
-        ["I am very disappointed with the service."],
-        ["It was okay, not bad."],
-        ["Absolutely fantastic experience!"]
-    ]
+print("✅ Model loaded successfully!")
+
+# تجربة بسيطة لتنبؤ
+test_sentence = "I love this movie!"
+tokens = tokenizer(
+    test_sentence,
+    padding='max_length',
+    truncation=True,
+    max_length=128,
+    return_tensors="np"
 )
 
-interface.launch()
+input_ids = tokens["input_ids"]
+prediction = model.predict(input_ids)[0][0]
+print(f"Prediction: {prediction}")