Why does the batch size significantly affect the recognition results in (iic/speech_paraformer-large-vad-punc_asr_nat-en-16k-common-vocab10020) and other models are not significantly affected by the batch size?paraformer-en长音频版对batch size（或者说padding操作）过于敏感，严重影响识别结果

[283258771]

Why does the batch size significantly affect the recognition results in this model and other models are not significantly affected by the batch size? 这个长音频版本的模型的识别结果对 batch size 特别敏感（主要是对 padding 操作很敏感），其它版本则没事...