Update utils/llm_xiapi.py

utils/llm_xiapi.py

@@ -16,7 +16,7 @@ The 'onset-offset' inside needs to be determined based on common sense, with a d
 You need to make a prediction for the total duration, which should not exceed 20 seconds and not exceed the latest end time of a single event.
 And pay attention to vocabulary that represents the order and frequency of events, such as 'after', 'followed by', 'n times', and so on.
 You can use the latest ending event of all events in the training dataset as the total audio time
-It is preferred that events do not overlap as much as possible.
+Events must not overlap in time under any circumstances.
 Now, I will provide you with some examples in training set for your learning, each example in the format 'index: input~output'. 
 {"onset": "squeal__1.359-2.373_3.216-4.23_5.576-6.59", "captions": "squeal 3 times", "length": "7.52"}
 {"onset": "sawing__1.432-3.975_4.533-6.54", "captions": "sawing 2 times", "length": "9.26"}