seaco_paraformer微调时报错：seaco_id = samples[0]["seaco_id"] IndexError: list index out of range

[1xuanxia]

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.

What is your question?

使用seaco_paraformer进行微调时，发现如下问题，已经使用最新的seaco_paraformer模型和finetune.sh代码，训练数据格式正确，仍然出现以下问题，funasr=1.2.0。

Code

File "/opt/asr_finetune/FunASR/examples/industrial_data_pretraining/seaco_paraformer/../../../funasr/bin/train_ds.py", line 56, in main_hydra
main(**kwargs)
File "/opt/asr_finetune/FunASR/examples/industrial_data_pretraining/seaco_paraformer/../../../funasr/bin/train_ds.py", line 173, in main
trainer.train_epoch(
File "/root/anaconda3/envs/funasr/lib/python3.10/site-packages/funasr/train_utils/trainer_ds.py", line 578, in train_epoch
for batch_idx, batch in enumerate(dataloader_train):
File "/root/anaconda3/envs/funasr/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 701, in next
data = self._next_data()
File "/root/anaconda3/envs/funasr/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 757, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "/root/anaconda3/envs/funasr/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 55, in fetch
return self.collate_fn(data)
File "/root/anaconda3/envs/funasr/lib/python3.10/site-packages/funasr/datasets/audio_datasets/datasets.py", line 215, in collator
seaco_id = samples[0]["seaco_id"]
IndexError: list index out of range
E0315 13:47:12.072000 618603 site-packages/torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 1) local_rank: 0 (pid: 618658) of binary: /root/anaconda3/envs/funasr/bin/python
Traceback (most recent call last):
File "/root/anaconda3/envs/funasr/bin/torchrun", line 33, in
sys.exit(load_entry_point('torch==2.5.1', 'console_scripts', 'torchrun')())

What have you tried?

What's your environment?

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

[zzhdbw]

same question

[Jadesandalwood]

same question~~

[Jadesandalwood]

same question~~

Modifying the configurations of

++dataset_conf.data_split_num=
++dataset_conf.batch_size=

can resolve this issue.

[R1310328554]

解决了一个，又出现另外一个问题， 没法做微调训练啊：

File "/www/ai/ocr/asr/FunASR/myFunASR/lib/python3.12/site-packages/torch/autograd/graph.py", line 841, in _engine_run_backward
[rank1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]: RuntimeError: Detected mismatch between collectives on ranks. Rank 1 is running collective: CollectiveFingerPrint(SequenceNumber=3124, OpType=ALLREDUCE, TensorShape=[4839636], TensorDtypes=Float, TensorDeviceTypes=TensorOptions(dtype=float (default), device=cuda, layout=Strided (default), requires_grad=false (default), pinned_memory=false (default), memory_format=(nullopt))), but Rank 0 is running collective: CollectiveFingerPrint(SequenceNumber=0OpType=ALLGATHER_COALESCED).Collectives differ in the following aspects: Sequence number: 3124vs 0 Op type: ALLREDUCEvs ALLGATHER_COALESCED Tensor Tensor shapes: 4839636vs Tensor Tensor dtypes: Floatvs Tensor Tensor devices: TensorOptions(dtype=float (default), device=cuda, layout=Strided (default), requires_grad=false (default), pinned_memory=false (default), memory_format=(nullopt))vs