Hugging Face's logo

Datasets:
BeardedMonster
/
data-gen

Tasks:
Text Generation
Languages:
English
Yoruba
Hausa
Tags:
west-african-languages
tool-calling
Synthetic
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    TypeError
Message:      Couldn't cast array of type
struct<pcm: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, yor: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, hau: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ibo: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, twi: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, aka: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, efi: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, urh: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, fon: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ewe: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ful: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, fuv: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, eng: struct<code: string, name: string, tier: string, samples: int64, guidance: string>>
to
{'pcm': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'yor': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'hau': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ibo': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'twi': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'aka': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'efi': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'urh': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'fon': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ewe': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ful': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'fuv': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}}
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2312, in cast_table_to_schema
                  cast_array_to_feature(
                  ~~~~~~~~~~~~~~~~~~~~~^
                      table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                      feature,
                      ^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
                  return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
                                           ~~~~^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2158, 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<pcm: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, yor: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, hau: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ibo: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, twi: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, aka: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, efi: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, urh: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, fon: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ewe: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, ful: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, fuv: struct<code: string, name: string, tier: string, samples: int64, guidance: string>, eng: struct<code: string, name: string, tier: string, samples: int64, guidance: string>>
              to
              {'pcm': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'yor': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'hau': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ibo': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'twi': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'aka': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'efi': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'urh': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'fon': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ewe': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'ful': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}, 'fuv': {'code': Value('string'), 'name': Value('string'), 'tier': Value('string'), 'samples': Value('int64'), 'guidance': Value('string')}}

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

SabiYarn data-gen

Synthetic training data for SabiYarn, a ~306M parameter MoE model for 12 West African languages plus English.

Design

The model is small, so this corpus does not try to pack world facts into its weights. It teaches general understanding of how things work, plus two reflexes: reason inside <think>...</think>, then either use a tool or say plainly that it does not know. A confident invention is the worst outcome; an honest "I don't know" is always preferred.

Splits

folder contents
pretrain/ plain prose documents, no markup, per language
sft/ 6-10 message conversations with tool calls, messages + rendered text
rl/ conversation prefix + 2-3 ranked candidate final replies
seeds/ the briefs that generated all of it -- read these first

pretrain

Target: 420,000 samples across 12 languages.

sft

Target: 185,000 samples across 13 languages.

task tags share
world_knowledge_qa qa, multi-turn-chat 11.0%
tool_search_answer tool-calling, knowledge-boundary, qa 9.0%
rag_document_qa rag, extractive-qa, tool-calling 8.0%
no_tool_admit_unknown knowledge-boundary, qa 6.5%
retrieval_insufficient insufficient-context, rag, tool-calling 6.5%
tool_compute tool-calling, structured-output 4.0%
tool_database tool-calling, structured-output 4.0%
action_tool_use tool-calling 4.0%
health_education health-education 4.0%
structured_output structured-output 4.0%
translation_english translation 4.0%
summarization summarization 4.0%
rephrasing_and_writing text-rephrasing, content-writing 4.0%
health_advice health-advice, tool-calling 3.5%
financial_analysis financial-analysis, tool-calling 3.5%
health_triage health-triaging 3.0%
translation_interlanguage translation 3.0%
general_chat multi-turn-chat 2.5%
topic_classification topic-classification 2.0%
sentiment_analysis sentiment-analysis 2.0%
intent_detection intent-detection 1.5%
toxicity_detection toxicity-spam-detection 1.5%
language_identification language-identification 1.5%
ner ner 1.5%
pos_tagging pos-tagging 1.5%

rl

Target: 65,000 samples across 13 languages.

task tags share
tool_search_answer tool-calling, knowledge-boundary, qa 15.0%
retrieval_insufficient insufficient-context, rag, tool-calling 14.0%
rag_document_qa rag, extractive-qa, tool-calling 13.0%
no_tool_admit_unknown knowledge-boundary, qa 12.0%
world_knowledge_qa qa, multi-turn-chat 9.0%
tool_compute tool-calling, structured-output 6.0%
action_tool_use tool-calling 6.0%
health_advice health-advice, tool-calling 6.0%
tool_database tool-calling, structured-output 5.0%
health_triage health-triaging 5.0%
financial_analysis financial-analysis, tool-calling 4.0%
structured_output structured-output 2.0%
translation_english translation 2.0%
translation_interlanguage translation 1.0%

Shards currently in this repo

split languages shards
sft 4 12

Provenance

Generated with data-gen/ in the SabiYarn repo, via Together AI / OpenRouter (openai/gpt-oss-120b and gemma-3-27b). Every sample carries the id that ties it back to a deterministic plan row in its seed file, so any sample can be traced to the exact brief, language, task and domain that produced it.

Synthetic data in low-resource languages is imperfect. Efik, Urhobo, Fon, Ewe and Fulfulde in particular should be spot-checked by speakers before being trusted.
Downloads last month
-
Total file size:
867 kB