IndexError: index 3 is out of bounds for dimension 1 with size 2

[JonneryR]

Question:
[rank0]: File "/home/jovyan/.conda/envs/funasr/lib/python3.11/site-packages/funasr/models/sense_voice/model.py", line 758, in encode
[rank0]: [[self.textnorm_int_dict[int(style)]] for style in text[:, 3]]
[rank0]: ~~~~^^^^^^
[rank0]: IndexError: index 3 is out of bounds for dimension 1 with size 2

What have you tried?

1.I try to print the shape of text:
torch.Size([82, 3])
torch.Size([89, 2])
2.I try to add text_language, emo and event in the jsonl file:
e.g
{"key": "wechat_msg_chunk_12135561624217528293_1709826135802_external_left_chunk_32_16k_real_8k_alaw__629597", "source": "/home/jovyan/jonneryr/asr/sales/data/train/label/train_wavs_wer_less_than_5_16k/wechat_msg_chunk_12135561624217528293_1709826135802_external_left_chunk_32_16k_real_8k_alaw_.wav", "source_len": 261, "target": "零六没有", "target_len": 4, "emo_target": "<|NEUTRAL|>", "emo_target_len": 11, "event_target": "<|Speech|>", "event_target_len": 10, "text_language": "<|zh|>", "text_language_len": 6}

What's your environment?

• OS (e.g., Linux):Linux
• FunASR Version (e.g., 1.0.0):1.1.14
• ModelScope Version (e.g., 1.11.0):1.19.2
• PyTorch Version (e.g., 2.0.0):2.5.1
• How you installed funasr (pip, source):pip
• Python version:3.11
• GPU (e.g., V100M32) A800
• CUDA/cuDNN version (e.g., cuda11.7):cuda12.4
• Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1)
• Any other relevant information:

[JonneryR]

看起来好像是speech 要做padding，text应该也要做一些padding才对？
但是没有这部分代码。

[JonneryR]

问题解决了，是dataset的选择问题，需要选择SenseVoiceCTCDataset，只有这里才有给text前面做padding的代码。

[ouyang982020]

问题解决了，是dataset的选择问题，需要选择SenseVoiceCTCDataset，只有这里才有给text前面做padding的代码。

请问一下，我这边也遇到了同样的问题，能告知是哪个位置去做修改+选择吗

[JonneryR]

就是训练的时候dataset 设置成 SenseVoiceCTCDataset

[ouyang982020]

就是训练的时候dataset 设置成 SenseVoiceCTCDataset

好的，我好像明白了，命令行里面可以加