TypeError: Couldn't cast array of type struct<answer: struct<unanswerable: bool, answerType: string, free_form_answer: string, evidence: list<item: string>, evidenceAnnotate: list<item: string>, highlighted_evidence: list<item: string>>> to ...

### Describe the bug

TypeError: Couldn't cast array of type struct, evidenceAnnotate: list, highlighted_evidence: list>> to {'answer': {'unanswerable': Value(dtype='bool', id=None), 'answerType': Value(dtype='string', id=None), 'free_form_answer': Value(dtype='string', id=None), 'evidence': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'evidenceAnnotate': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'highlighted_evidence': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None)}, 'unanswerable': Value(dtype='bool', id=None), 'answerType': Value(dtype='string', id=None), 'free_form_answer': Value(dtype='string', id=None), 'evidence': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'evidenceAnnotate': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'highlighted_evidence': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None)}

When I use _load_dataset()_ I get the error

`from datasets import load_dataset
datafiles = {'train': './data/train.json', 'validation': './data/validation.json', 'test': './data/test.json'}
raw_data = load_dataset("json", data_files=datafiles, cache_dir="./cache")
`
Detailed error information is as follows：

Traceback (most recent call last):
File "C:/Users/CHENJIALEI/Desktop/NLPCC2023/NLPCC23_SciMRC-main/test2.py", line 9, in
raw_data = load_dataset("json", data_files=datafiles, cache_dir="./cache")
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\load.py", line 1747, in load_dataset
builder_instance.download_and_prepare(
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\builder.py", line 814, in download_and_prepare
self._download_and_prepare(
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\builder.py", line 905, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\builder.py", line 1521, in _prepare_split
writer.write_table(table)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\arrow_writer.py", line 540, in write_table
pa_table = table_cast(pa_table, self._schema)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 2069, in table_cast
return cast_table_to_schema(table, schema)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 2031, in cast_table_to_schema
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 2031, in
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1740, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1740, in
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1867, in cast_array_to_feature
casted_values = _c(array.values, feature[0])
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1742, in wrapper
return func(array, *args, **kwargs)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1862, in cast_array_to_feature
arrays = [_c(array.field(name), subfeature) for name, subfeature in feature.items()]
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1862, in
arrays = [_c(array.field(name), subfeature) for name, subfeature in feature.items()]
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1742, in wrapper
return func(array, *args, **kwargs)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1867, in cast_array_to_feature
casted_values = _c(array.values, feature[0])
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1742, in wrapper
return func(array, *args, **kwargs)
File "D:\Environment\anaconda3\envs\test\lib\site-packages\datasets\table.py", line 1913, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{array.type}\nto\n{feature}")

It is successful when I load the data separately

`raw_data = load_dataset("json", data_files="./data/train.json", cache_dir="./cache")`

### Steps to reproduce the bug

1.from datasets import load_dataset
2.datafiles = {'train': './data/train.json', 'validation': './data/validation.json', 'test': './data/test.json'}
3.raw_data = load_dataset("json", data_files=datafiles, cache_dir="./cache")

### Expected behavior

Successfully load dataset

### Environment info

datasets == 2.6.1
pyarrow == 8.0.0
python == 3.8

platform:windows11