Add float(), reuqired for CPU

app.py

@@ -44,7 +44,7 @@ if has_cuda:
 else:
     model = AutoModel.from_pretrained(
         model_name, trust_remote_code=True
-    )  # .float() .half().float()
+    ).half().float()  # .float() .half().float(): musr use float for cpu
 
 model = model.eval()
 logger.debug("done load")