Set temperature to lowest value for evaluation script

eval_prompt.py

@@ -23,6 +23,8 @@ def evaluate_prompt(prompt: str, task_name: str, args: Namespace):
         load_in_8bit=True,
         torch_dtype=torch.float16,
         device_map="auto",
+        # use less randomness, i.e., more certain outputs
+        temperature=0.0,
         # torch_dtype is not JSON serializable therefore we ignore it
         ignore_cache_kwargs=["torch_dtype"],
     )