minor update

Author: Liyang90

llama/generation.py

@@ -74,8 +74,8 @@ def generate(
 
         # Passing tensors instead of floats into self._generate_one_token_fn,
         # so that different values would not trigger compilations of new graphs
-        temperature_tensor = torch.tensor(temperature).to(device)
-        top_p_tensor = torch.tensor(top_p).to(device)
+        temperature_tensor = torch.tensor(float(temperature)).to(device)
+        top_p_tensor = torch.tensor(float(top_p)).to(device)
         with_temp = temperature > 0
 
         cache_kvs = self.model.cache_kvs