aaron-lii

We can change some words in one audio to other words by using this function. ## Here is an example: Original audio text is "XTTS is a Voice generation model...

1. 根据flow推理代码对训练时的forward部分进行了修正。 2. 根据example.py中的示例，增加了llm训练时instruct_token的拼接，与推理结构保持一致。 改动在单卡训练上已跑通。