deploy

app.py

@@ -20,7 +20,12 @@ app = FastAPI()
 model_name = "ihanif/whisper-medium-urdu"
 try:
     logger.info(f"Loading processor for {model_name}")
-    processor = WhisperProcessor.from_pretrained(model_name, language="urdu", task="transcribe")
+    processor = WhisperProcessor.from_pretrained(
+        model_name,
+        language="urdu",
+        task="transcribe",
+        clean_up_tokenization_spaces=True  # Suppress FutureWarning
+    )
     logger.info(f"Loading model for {model_name}")
     model = WhisperForConditionalGeneration.from_pretrained(model_name, low_cpu_mem_usage=True)
 except Exception as e:
@@ -84,11 +89,10 @@ async def transcribe_audio(file: UploadFile = File(...)):
             return generated_ids
 
         try:
-            async with asyncio.timeout(60):  # 60-second timeout
-                generated_ids = await generate_transcription()
+            generated_ids = await asyncio.wait_for(generate_transcription(), timeout=60)  # 60-second timeout
         except asyncio.TimeoutError:
             logger.error("Transcription timed out after 60 seconds")
-            return {"error": "Transcription took too long. Try a smaller model or upgrade your Space."}
+            return {"error": "Transcription took too long. Try a smaller model (e.g., whisper-small) or upgrade to a paid Hugging Face Space with GPU."}
 
         transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
         total_time = time.time() - start_time