Update app.py

app.py

@@ -9,22 +9,24 @@ import os
 model = None
 processor = None
 
-print("Loading model...")
-model, _ = FastModel.from_pretrained(
-    model_name = "oddadmix/gemma-4b-egyptian-code-switching-b4-g2",
-    dtype = None,
-    max_seq_length = 2048,
-    load_in_4bit = True,  # Enable 4bit for GPU memory efficiency
-    full_finetuning = False,
-)
-
-processor = Gemma3nProcessor.from_pretrained("google/gemma-3n-E4B-it")
-
-# Set model to inference mode
-FastLanguageModel.for_inference(model)
-print("Model loaded successfully!")
-
-
+def load_model():
+    """Load the model and processor once at startup"""
+    global model, processor
+    
+    print("Loading model...")
+    model, _ = FastModel.from_pretrained(
+        model_name = "oddadmix/gemma-4b-egyptian-code-switching-b4-g2",
+        dtype = None,
+        max_seq_length = 2048,
+        load_in_4bit = True,  # Enable 4bit for GPU memory efficiency
+        full_finetuning = False,
+    )
+    
+    processor = Gemma3nProcessor.from_pretrained("google/gemma-3n-E4B-it")
+    
+    # Set model to inference mode
+    FastLanguageModel.for_inference(model)
+    print("Model loaded successfully!")
 @spaces.GPU
 def transcribe_audio(audio_path, max_tokens=128):
     """Transcribe audio file using the loaded model"""
@@ -140,4 +142,5 @@ with gr.Blocks(title="Egyptian Arabic ASR") as demo:
 
 # Launch the app
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()
+