Update handler.py

handler.py

@@ -3,7 +3,7 @@ from tangoflux import TangoFluxInference
 import torchaudio
 
 from huggingface_inference_toolkit.logging import logger
-
+import io
 
 class EndpointHandler():
     def __init__(self, path=""):
@@ -40,7 +40,12 @@ class EndpointHandler():
         duration = parameters.get("duration", 10)
         guidance_scale = parameters.get("guidance_scale", 3.5)
 
-        
-        return self.model.generate(prompt,steps=num_inference_steps,
+        audio= self.model.generate(prompt,steps=num_inference_steps,
         duration=duration,
-        guidance_scale=guidance_scale)
+        guidance_scale=guidance_scale)
+        
+        buffer = io.BytesIO()
+        torchaudio.save(buffer, audio, 44100, format="wav")
+        buffer.seek(0)
+        
+        return buffer.read()