Update app.py

app.py

@@ -1,3 +1,7 @@
+import gradio as gr
+from transformers import pipeline
+import os
+import numpy as np
 import torch
 import spaces
 
@@ -8,42 +12,42 @@ classifier = pipeline("audio-classification", model=model_id, device='cuda')
 print("Model loaded successfully")
 print("Model moved to GPU successfully")
 
-
 @spaces.GPU
 def predict(audio_segment, sr=16000):
     return classifier({"sampling_rate": sr, "raw": audio_segment})
 
 # define dialect mapping
-
 dialect_mapping = {
     "MSA": "Modern Standard Arabic (MSA) - العربية الفصحى الحديثة",
     "Egyptian": "Egyptian Arabic -  اللهجة المصرية العامية",
+    "Gulf": "Peninsular Arabic - لهجة الجزيرة العربية",
+    "Levantine": "Levantine Arabic - لهجة بلاد الشام",
+    "Maghrebi": "Maghrebi Arabic - اللهجة المغاربية"
+}
+
+def predict_dialect(audio):
     if audio is None:
         return {"Error": 1.0}
     
-
     sr, audio_array = audio
     
-
     if len(audio_array.shape) > 1:
         audio_array = audio_array.mean(axis=1)
 
-
     if audio_array.dtype != np.float32:
-
         if audio_array.dtype == np.int16:
             audio_array = audio_array.astype(np.float32) / 32768.0
         else:
+            audio_array = audio_array.astype(np.float32)
     
     print(f"Processing audio: sample rate={sr}, shape={audio_array.shape}")
     
-
     predictions = predict(sr=sr, audio_segment=audio_array)
     
-
     results = {}
     for pred in predictions:
         dialect_name = dialect_mapping.get(pred['label'], pred['label'])
+        results[dialect_name] = float(pred['score'])
     
     return results
 
@@ -54,14 +58,13 @@ if os.path.exists(examples_dir):
     for filename in os.listdir(examples_dir):
         if filename.endswith((".wav", ".mp3", ".ogg")):
             examples.append([os.path.join(examples_dir, filename)])
-
     print(f"Found {len(examples)} example files")
 else:
     print("Examples directory not found")
 
 # clean description without problematic HTML
 description = """
-By <a href="https://badrex.github.io/">Badr Alabsi</a> with ❤️🤍💚
+Developed with ❤️🤍💚 by <a href="https://badrex.github.io/">Badr al-Absi</a> 
 
 This is a demo for the accurate and robust Transformer-based <a href="https://huggingface.co/badrex/mms-300m-arabic-dialect-identifier">model</a> for Spoken Arabic Dialect Identification (ADI). 
 From just a short audio clip (5-10 seconds), the model can identify Modern Standard Arabic (MSA) as well as four major regional Arabic varieties: Egyptian Arabic, Peninsular Arabic (Gulf, Yemeni, and Iraqi), Levantine Arabic, and Maghrebi Arabic.
@@ -69,7 +72,6 @@ From just a short audio clip (5-10 seconds), the model can identify Modern Stand
 Simply **upload an audio file** 📤 or **record yourself speaking** 🎙️⏺️ to try out the model!
 """
 
-
 demo = gr.Interface(
     fn=predict_dialect,
     inputs=gr.Audio(),
@@ -78,9 +80,7 @@ demo = gr.Interface(
     description=description,
     examples=examples if examples else None,
     cache_examples=False,
-
     flagging_mode=None
 )
 
-
 demo.launch(share=True)