Dataset Viewer issue: TypeError: Couldn't cast array
#5
by
albertvillanova
- opened
The dataset viewer is not working.
Error details:
Error code: DatasetGenerationError
Exception: TypeError
Message: Couldn't cast array of type
struct<identifier: int64, comment: string, is_minor_edit: bool, editor: struct<identifier: int64, name: string, is_anonymous: bool, edit_count: int64, groups: list<item: string>, is_patroller: bool, date_started: timestamp[s], is_admin: bool>, number_of_characters: int64, size: struct<value: int64, unit_text: string>, tags: list<item: string>, scores: struct<revertrisk: struct<probability: struct<false: double, true: double>, prediction: bool>>, maintenance_tags: struct<>, noindex: bool>
to
{'identifier': Value(dtype='int64', id=None), 'comment': Value(dtype='string', id=None), 'is_minor_edit': Value(dtype='bool', id=None), 'scores': {'revertrisk': {'probability': {'false': Value(dtype='float64', id=None), 'true': Value(dtype='float64', id=None)}, 'prediction': Value(dtype='bool', id=None)}}, 'editor': {'identifier': Value(dtype='int64', id=None), 'name': Value(dtype='string', id=None), 'edit_count': Value(dtype='int64', id=None), 'groups': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'date_started': Value(dtype='timestamp[s]', id=None), 'is_patroller': Value(dtype='bool', id=None), 'is_bot': Value(dtype='bool', id=None), 'is_admin': Value(dtype='bool', id=None), 'is_anonymous': Value(dtype='bool', id=None), 'has_advanced_rights': Value(dtype='bool', id=None)}, 'number_of_characters': Value(dtype='int64', id=None), 'size': {'value': Value(dtype='int64', id=None), 'unit_text': Value(dtype='string', id=None)}, 'noindex': Value(dtype='bool', id=None), 'maintenance_tags': {'pov_count': Value(dtype='int64', id=None), 'update_count': Value(dtype='int64', id=None), 'citation_needed_count': Value(dtype='int64', id=None)}, 'tags': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'is_breaking_news': Value(dtype='bool', id=None)}
Traceback: Traceback (most recent call last):
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2013, in _prepare_split_single
writer.write_table(table)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/arrow_writer.py", line 585, in write_table
pa_table = table_cast(pa_table, self._schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2302, in table_cast
return cast_table_to_schema(table, schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2261, in cast_table_to_schema
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2261, in <listcomp>
arrays = [cast_array_to_feature(table[name], feature) for name, feature in features.items()]
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 1802, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 1802, in <listcomp>
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2122, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
struct<identifier: int64, comment: string, is_minor_edit: bool, editor: struct<identifier: int64, name: string, is_anonymous: bool, edit_count: int64, groups: list<item: string>, is_patroller: bool, date_started: timestamp[s], is_admin: bool>, number_of_characters: int64, size: struct<value: int64, unit_text: string>, tags: list<item: string>, scores: struct<revertrisk: struct<probability: struct<false: double, true: double>, prediction: bool>>, maintenance_tags: struct<>, noindex: bool>
to
{'identifier': Value(dtype='int64', id=None), 'comment': Value(dtype='string', id=None), 'is_minor_edit': Value(dtype='bool', id=None), 'scores': {'revertrisk': {'probability': {'false': Value(dtype='float64', id=None), 'true': Value(dtype='float64', id=None)}, 'prediction': Value(dtype='bool', id=None)}}, 'editor': {'identifier': Value(dtype='int64', id=None), 'name': Value(dtype='string', id=None), 'edit_count': Value(dtype='int64', id=None), 'groups': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'date_started': Value(dtype='timestamp[s]', id=None), 'is_patroller': Value(dtype='bool', id=None), 'is_bot': Value(dtype='bool', id=None), 'is_admin': Value(dtype='bool', id=None), 'is_anonymous': Value(dtype='bool', id=None), 'has_advanced_rights': Value(dtype='bool', id=None)}, 'number_of_characters': Value(dtype='int64', id=None), 'size': {'value': Value(dtype='int64', id=None), 'unit_text': Value(dtype='string', id=None)}, 'noindex': Value(dtype='bool', id=None), 'maintenance_tags': {'pov_count': Value(dtype='int64', id=None), 'update_count': Value(dtype='int64', id=None), 'citation_needed_count': Value(dtype='int64', id=None)}, 'tags': Sequence(feature=Value(dtype='string', id=None), length=-1, id=None), 'is_breaking_news': Value(dtype='bool', id=None)}
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1391, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 990, in stream_convert_to_parquet
builder._prepare_split(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1884, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2040, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset
I am investigating it.
We will need to define the data types explicitly and avoid their inference.
However, we also need to fix some underlying issue in the datasets library with JSON-lines data files. I opened PRs:
Thanks for all the help
@albertvillanova
!
Looks like the issues are fixed on your end?
We are having a closer look at the schema updates on our end and will get back to you.
Hello,
I have the same issue when loading the dataset:
I have downloaded and unzipped the dataset locally but when loading I have the same issue.
import datasets
dataset = datasets.load_dataset("/gpfsdsdir/dataset/HuggingFace/wikimedia/structured-wikipedia/20240916.fr")
TypeError: Couldn't cast array of type
struct<content_url: string, width: int64, height: int64, alternative_text: string>
to
{'content_url': Value(dtype='string', id=None), 'width': Value(dtype='int64', id=None), 'height': Value(dtype='int64', id=None)}
The above exception was the direct cause of the following exception:
My version of datasets is 3.0.1
Thanks for reporting @Aremaki ,
You are using the right datasets library version: the required fixes to support missing fields were released in datasets-3.0.1: https://github.com/huggingface/datasets/releases/tag/3.0.1
However the issue here is not a missing field, but a present field which is not defined in the README schema:
- a data item with 4 fields was found:
- content_url: string
- width: int64
- height: int64
- alternative_text: string
- however the expected schema contains only 3 fields:
- content_url: Value(dtype='string', id=None)
- width: Value(dtype='int64', id=None)
- height: Value(dtype='int64', id=None)
While trying to load the dataset, I discovered several schema misalignments between the expected (either provided in the README or inferred) schema and the real data:
- sections.has_parts.has_parts.has_parts.has_parts.name
- sections.has_parts.has_parts.has_parts.has_parts.has_parts.links.images
- sections.has_parts.has_parts.has_parts.has_parts.has_parts.has_parts
- sections.has_parts.has_parts.has_parts.has_parts.has_parts.has_parts.links
I reported them to the Wikimedia team. And they replied they are working on updating their schema to align it with the data.
Thanks
@Aremaki
and
@albertvillanova
!
Happy to report the work on schema updates is part of our current sprint, you can follow the ticket here: https://phabricator.wikimedia.org/T375462. We'll let you know when this is completed.