share true

app.py

@@ -4,19 +4,19 @@ import torch
 from transformers import pipeline
 import spaces
 
+# Initialize model once at startup
+pipe = pipeline(
+    model="sarvamai/shuka_v1",
+    trust_remote_code=True,
+    device=0 if torch.cuda.is_available() else -1
+)
+
 @spaces.GPU
 def transcribe_and_respond(audio_file):
     try:
         # Load audio
         audio, sr = librosa.load(audio_file, sr=16000)
 
-        # Load Shuka model via pipeline
-        pipe = pipeline(
-            model="sarvamai/shuka_v1",
-            trust_remote_code=True,
-            device=0 if torch.cuda.is_available() else -1
-        )
-
         # Use Shuka's expected format
         output = pipe({
             "audio": audio,