Update app.py

app.py

@@ -11,7 +11,9 @@ model_id = "sesame/csm-1b"
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 processor = AutoProcessor.from_pretrained(model_id, token=HF_TOKEN)
-model = CsmForConditionalGeneration.from_pretrained(model_id, device_map=device, token=HF_TOKEN)
+
+model = CsmForConditionalGeneration.from_pretrained(model_id, token=HF_TOKEN)
+model.to(device)
 
 app = FastAPI()