paraformer微调之后模型变大，且和basemodel推理同一段wav文件时会报错

[YouTwoMeToo]

在对paraformer长音频版模型进行微调之后，保存的pt文件大小由basemodel的800多M增加到了近2.6G，
且在推理同一段wav文件时，会报错，报错信息如下：

Traceback (most recent call last):
File "/wind/aispace/train/source/src/FunASR/examples/industrial_data_pretraining/paraformer-zh-spk/tasks_bin.py", line 220, in
results_left = asr_batch_infer(output_left_folder,paraformer_model)
File "/wind/aispace/train/source/src/FunASR/examples/industrial_data_pretraining/paraformer-zh-spk/tasks_bin.py", line 124, in asr_batch_infer
res = paraformer_model.generate(input=audio_binary,fs=8000)
File "/wind/aispace/train/source/src/FunASR/funasr/auto/auto_model.py", line 300, in generate
return self.inference(input, input_len=input_len, **cfg)
File "/wind/aispace/train/source/src/FunASR/funasr/auto/auto_model.py", line 342, in inference
res = model.inference(**batch, **kwargs)
File "/wind/aispace/train/source/src/FunASR/funasr/models/bicif_paraformer/model.py", line 351, in inference
postprocess_utils.sentence_postprocess(token, timestamp)
File "/wind/aispace/train/source/src/FunASR/funasr/utils/postprocess_utils.py", line 235, in sentence_postprocess
word_lists, ts_lists = abbr_dispose(word_lists, ts_lists)
File "/wind/aispace/train/source/src/FunASR/funasr/utils/postprocess_utils.py", line 131, in abbr_dispose
begin = time_stamp[ts_nums[num]][0]
IndexError: list index out of range
0%|

funasr为最新版
请问这个问题是什么原因呢？会是与微调的数据有关系吗？

[YouTwoMeToo]

您好，今天又测试了一下，在训练数据中加入了比较短的片段，效果有改善，但是依然会有少部分测试用例报了如上错误