为什么使用paraformer-large模型的长音频版运行finetune.sh脚本，还是无法识别20s以上的音频文件

[lllmd]

Notice: In order to resolve issues more efficiently, please raise issue following the template.
（注意：为了更加高效率解决您遇到的问题，请按照模板提问，补充细节）

❓ Questions and Help

Before asking:

1. search the issues.
2. search the docs.

之前使用paraformer-large模型，在finetune.sh文件中添加了max_token_length参数后，仍然无法识别大于20s的音频文件，在更换了paraformer-large的长音频版本后，还是出现同样的问题。

这是运行时显示的内容：
{'scp_file_list': ['/home/ubuntu1/data/list/train_wav.scp', '/home/ubuntu1/data/list/train_text.txt'], 'data_type_list': ['source', 'target'], 'jsonl_file_out': '/home/ubuntu1/data/list/train.jsonl'}
convert wav.scp text to jsonl, ncpu: 32
cpu: 0: 100%|██████████████████████████████████████████████████████████████████████| 5/5 [00:00<00:00, 5.67it/s]
cpu: 0: 100%|████████████████████████████████████████████████████████████████████| 5/5 [00:00<00:00, 4804.47it/s]
processed 5 samples
{'scp_file_list': ['/home/ubuntu1/data/list/val_wav.scp', '/home/ubuntu1/data/list/val_text.txt'], 'data_type_list': ['source', 'target'], 'jsonl_file_out': '/home/ubuntu1/data/list/val.jsonl'}
convert wav.scp text to jsonl, ncpu: 32
cpu: 0: 100%|██████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 2.29it/s]
cpu: 0: 100%|████████████████████████████████████████████████████████████████████| 2/2 [00:00<00:00, 3818.21it/s]
processed 2 samples
log_file: ./outputs/log.txt

这是log文件中显示的内容：
Model summary:
Class Name: BiCifParaformer
Total Number of model parameters: 225.07 M
Number of trainable parameters: 225.07 M (100.0%)
Type: torch.float32
[2024-06-24 15:32:07,818][root][INFO] - Build optim
[2024-06-24 15:32:07,822][root][INFO] - Build scheduler
[2024-06-24 15:32:07,823][root][INFO] - Build dataloader
[2024-06-24 15:32:07,823][root][INFO] - Build dataloader
[2024-06-24 15:32:07,823][root][INFO] - total_num of samplers: 1, /home/ubuntu1/data/list/train.jsonl
[2024-06-24 15:32:07,823][root][INFO] - total_num of samplers: 2, /home/ubuntu1/data/list/val.jsonl
[2024-06-24 15:32:07,823][root][WARNING] - distributed is not initialized, only single shard
[2024-06-24 15:32:07,853][root][INFO] - Train epoch: 0, rank: 0

What have you tried?

在finetune.sh中已经添加：
++dataset_conf.max_token_length=30000
但是还是没有作用

What's your environment?

• OS (e.g., Linux):
• FunASR Version (e.g., 1.0.27):
• ModelScope Version (e.g., 1.15.0):
• PyTorch Version (e.g., 2.3.0):
• How you installed funasr (pip, source):
• Python version:
• GPU (e.g., Tesla P40)
• CUDA/cuDNN version (e.g., cuda12.4):
• Any other relevant information:

[LauraGPT]

https://github.com/modelscope/FunASR/blob/main/funasr/datasets/audio_datasets/index_ds.py#L20