Datasets:

Tetsuto
/

pocketpedia-index

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
format: int64
tier: string
source: string
articles: int64
build_options: struct<no_leads: bool, zstd_dict: bool, strip_tail: bool, max_words: int64, full_text: bool, profile (... 25 chars omitted)
  child 0, no_leads: bool
  child 1, zstd_dict: bool
  child 2, strip_tail: bool
  child 3, max_words: int64
  child 4, full_text: bool
  child 5, profile: string
  child 6, version: int64
files: list<item: struct<name: string, size: int64, sha256: string>>
  child 0, item: struct<name: string, size: int64, sha256: string>
      child 0, name: string
      child 1, size: int64
      child 2, sha256: string
snapshot: struct<from: timestamp[s], to: timestamp[s]>
  child 0, from: timestamp[s]
  child 1, to: timestamp[s]
original_text_bytes: int64
options: struct<no_leads: bool, zstd_dict: bool, strip_tail: bool, max_words: int64, full_text: bool, profile (... 25 chars omitted)
  child 0, no_leads: bool
  child 1, zstd_dict: bool
  child 2, strip_tail: bool
  child 3, max_words: int64
  child 4, full_text: bool
  child 5, profile: string
  child 6, version: int64
dictionary: struct<seed: int64, articles: int64, eligible_articles: int64, sample_bytes: int64, bytes_per_articl (... 60 chars omitted)
  child 0, seed: int64
  child 1, articles: int64
  child 2, eligible_articles: int64
  child 3, sample_bytes: int64
  child 4, bytes_per_article_limit: int64
  child 5, requested_bytes: int64
  child 6, actual_bytes: int64
stored_text_bytes: int64
tail_cut_articles: int64
derived_lead_changes:
...
hild 3, max_words: int64
  child 7, gap_rule: string
  child 8, counts: struct<abstract_fallback: int64, articles: int64, empty_first_infoboxes: int64, empty_text_skipped:  (... 729 chars omitted)
      child 0, abstract_fallback: int64
      child 1, articles: int64
      child 2, empty_first_infoboxes: int64
      child 3, empty_text_skipped: int64
      child 4, extra_infoboxes_not_rendered: int64
      child 5, gap_sample_articles: int64
      child 6, gap_sample_prose_hits: int64
      child 7, gap_sample_text_hits: int64
      child 8, infobox_after_body: int64
      child 9, infobox_after_second_line: int64
      child 10, infobox_cut_by_word_cap: int64
      child 11, infobox_in_derived_lead: int64
      child 12, infobox_survived_word_cap: int64
      child 13, infoboxes_rendered: int64
      child 14, nested_table_cells_not_rendered: int64
      child 15, table_head_repeats: int64
      child 16, table_references_missing: int64
      child 17, table_rows_cut_by_row_cap: int64
      child 18, table_rows_cut_by_word_cap: int64
      child 19, table_rows_rendered: int64
      child 20, table_rows_survived_word_cap: int64
      child 21, tables_capped: int64
      child 22, tables_in_skipped_sections: int64
      child 23, tables_rendered: int64
      child 24, tables_without_rows_skipped: int64
      child 25, tail_like_lines_skipped: int64
      child 26, tail_sections_skipped: int64
build_seconds: double
docs: int64
word_removed_chars: int64
tail_removed_chars: int64
to
{'options': {'no_leads': Value('bool'), 'zstd_dict': Value('bool'), 'strip_tail': Value('bool'), 'max_words': Value('int64'), 'full_text': Value('bool'), 'profile': Value('string'), 'version': Value('int64')}, 'tail_cut_articles': Value('int64'), 'word_cut_articles': Value('int64'), 'tail_removed_chars': Value('int64'), 'word_removed_chars': Value('int64'), 'derived_lead_changes': Value('int64'), 'original_text_bytes': Value('int64'), 'stored_text_bytes': Value('int64'), 'dictionary': {'seed': Value('int64'), 'articles': Value('int64'), 'eligible_articles': Value('int64'), 'sample_bytes': Value('int64'), 'bytes_per_article_limit': Value('int64'), 'requested_bytes': Value('int64'), 'actual_bytes': Value('int64')}, 'structured': {'source': Value('string'), 'snapshot': {'from': Value('timestamp[s]'), 'to': Value('timestamp[s]')}, 'shards': Value('int64'), 'row_groups': Value('int64'), 'sample_every_kth_row_group': Value('null'), 'dedupe': {'rows': Value('int64'), 'unique_pages': Value('int64'), 'duplicates_dropped': Value('int64')}, 'caps': {'table_max_rows': Value('int64'), 'passage_words': List(Value('int64')), 'head_repeat_max_words': Value('int64'), 'max_words': Value('int64')}, 'gap_rule': Value('string'), 'counts': {'abstract_fallback': Value('int64'), 'articles': Value('int64'), 'empty_first_infoboxes': Value('int64'), 'empty_text_skipped': Value('int64'), 'extra_infoboxes_not_rendered': Value('int64'), 'gap_sample_articles': Value('int64'), 'gap_sample_prose_hits': Value('int64'), 'gap_sample_text_hits': Value('int64'), 'infobox_after_body': Value('int64'), 'infobox_after_second_line': Value('int64'), 'infobox_cut_by_word_cap': Value('int64'), 'infobox_in_derived_lead': Value('int64'), 'infobox_survived_word_cap': Value('int64'), 'infoboxes_rendered': Value('int64'), 'nested_table_cells_not_rendered': Value('int64'), 'table_head_repeats': Value('int64'), 'table_references_missing': Value('int64'), 'table_rows_cut_by_row_cap': Value('int64'), 'table_rows_cut_by_word_cap': Value('int64'), 'table_rows_rendered': Value('int64'), 'table_rows_survived_word_cap': Value('int64'), 'tables_capped': Value('int64'), 'tables_in_skipped_sections': Value('int64'), 'tables_rendered': Value('int64'), 'tables_without_rows_skipped': Value('int64'), 'tail_like_lines_skipped': Value('int64'), 'tail_sections_skipped': Value('int64')}}, 'docs': Value('int64'), 'build_seconds': Value('float64'), 'vacuum': {'before_bytes': Value('int64'), 'after_bytes': Value('int64'), 'seconds': Value('float64')}}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, 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 129, 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 544, 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 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, 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 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              format: int64
              tier: string
              source: string
              articles: int64
              build_options: struct<no_leads: bool, zstd_dict: bool, strip_tail: bool, max_words: int64, full_text: bool, profile (... 25 chars omitted)
                child 0, no_leads: bool
                child 1, zstd_dict: bool
                child 2, strip_tail: bool
                child 3, max_words: int64
                child 4, full_text: bool
                child 5, profile: string
                child 6, version: int64
              files: list<item: struct<name: string, size: int64, sha256: string>>
                child 0, item: struct<name: string, size: int64, sha256: string>
                    child 0, name: string
                    child 1, size: int64
                    child 2, sha256: string
              snapshot: struct<from: timestamp[s], to: timestamp[s]>
                child 0, from: timestamp[s]
                child 1, to: timestamp[s]
              original_text_bytes: int64
              options: struct<no_leads: bool, zstd_dict: bool, strip_tail: bool, max_words: int64, full_text: bool, profile (... 25 chars omitted)
                child 0, no_leads: bool
                child 1, zstd_dict: bool
                child 2, strip_tail: bool
                child 3, max_words: int64
                child 4, full_text: bool
                child 5, profile: string
                child 6, version: int64
              dictionary: struct<seed: int64, articles: int64, eligible_articles: int64, sample_bytes: int64, bytes_per_articl (... 60 chars omitted)
                child 0, seed: int64
                child 1, articles: int64
                child 2, eligible_articles: int64
                child 3, sample_bytes: int64
                child 4, bytes_per_article_limit: int64
                child 5, requested_bytes: int64
                child 6, actual_bytes: int64
              stored_text_bytes: int64
              tail_cut_articles: int64
              derived_lead_changes:
              ...
              hild 3, max_words: int64
                child 7, gap_rule: string
                child 8, counts: struct<abstract_fallback: int64, articles: int64, empty_first_infoboxes: int64, empty_text_skipped:  (... 729 chars omitted)
                    child 0, abstract_fallback: int64
                    child 1, articles: int64
                    child 2, empty_first_infoboxes: int64
                    child 3, empty_text_skipped: int64
                    child 4, extra_infoboxes_not_rendered: int64
                    child 5, gap_sample_articles: int64
                    child 6, gap_sample_prose_hits: int64
                    child 7, gap_sample_text_hits: int64
                    child 8, infobox_after_body: int64
                    child 9, infobox_after_second_line: int64
                    child 10, infobox_cut_by_word_cap: int64
                    child 11, infobox_in_derived_lead: int64
                    child 12, infobox_survived_word_cap: int64
                    child 13, infoboxes_rendered: int64
                    child 14, nested_table_cells_not_rendered: int64
                    child 15, table_head_repeats: int64
                    child 16, table_references_missing: int64
                    child 17, table_rows_cut_by_row_cap: int64
                    child 18, table_rows_cut_by_word_cap: int64
                    child 19, table_rows_rendered: int64
                    child 20, table_rows_survived_word_cap: int64
                    child 21, tables_capped: int64
                    child 22, tables_in_skipped_sections: int64
                    child 23, tables_rendered: int64
                    child 24, tables_without_rows_skipped: int64
                    child 25, tail_like_lines_skipped: int64
                    child 26, tail_sections_skipped: int64
              build_seconds: double
              docs: int64
              word_removed_chars: int64
              tail_removed_chars: int64
              to
              {'options': {'no_leads': Value('bool'), 'zstd_dict': Value('bool'), 'strip_tail': Value('bool'), 'max_words': Value('int64'), 'full_text': Value('bool'), 'profile': Value('string'), 'version': Value('int64')}, 'tail_cut_articles': Value('int64'), 'word_cut_articles': Value('int64'), 'tail_removed_chars': Value('int64'), 'word_removed_chars': Value('int64'), 'derived_lead_changes': Value('int64'), 'original_text_bytes': Value('int64'), 'stored_text_bytes': Value('int64'), 'dictionary': {'seed': Value('int64'), 'articles': Value('int64'), 'eligible_articles': Value('int64'), 'sample_bytes': Value('int64'), 'bytes_per_article_limit': Value('int64'), 'requested_bytes': Value('int64'), 'actual_bytes': Value('int64')}, 'structured': {'source': Value('string'), 'snapshot': {'from': Value('timestamp[s]'), 'to': Value('timestamp[s]')}, 'shards': Value('int64'), 'row_groups': Value('int64'), 'sample_every_kth_row_group': Value('null'), 'dedupe': {'rows': Value('int64'), 'unique_pages': Value('int64'), 'duplicates_dropped': Value('int64')}, 'caps': {'table_max_rows': Value('int64'), 'passage_words': List(Value('int64')), 'head_repeat_max_words': Value('int64'), 'max_words': Value('int64')}, 'gap_rule': Value('string'), 'counts': {'abstract_fallback': Value('int64'), 'articles': Value('int64'), 'empty_first_infoboxes': Value('int64'), 'empty_text_skipped': Value('int64'), 'extra_infoboxes_not_rendered': Value('int64'), 'gap_sample_articles': Value('int64'), 'gap_sample_prose_hits': Value('int64'), 'gap_sample_text_hits': Value('int64'), 'infobox_after_body': Value('int64'), 'infobox_after_second_line': Value('int64'), 'infobox_cut_by_word_cap': Value('int64'), 'infobox_in_derived_lead': Value('int64'), 'infobox_survived_word_cap': Value('int64'), 'infoboxes_rendered': Value('int64'), 'nested_table_cells_not_rendered': Value('int64'), 'table_head_repeats': Value('int64'), 'table_references_missing': Value('int64'), 'table_rows_cut_by_row_cap': Value('int64'), 'table_rows_cut_by_word_cap': Value('int64'), 'table_rows_rendered': Value('int64'), 'table_rows_survived_word_cap': Value('int64'), 'tables_capped': Value('int64'), 'tables_in_skipped_sections': Value('int64'), 'tables_rendered': Value('int64'), 'tables_without_rows_skipped': Value('int64'), 'tail_like_lines_skipped': Value('int64'), 'tail_sections_skipped': Value('int64')}}, 'docs': Value('int64'), 'build_seconds': Value('float64'), 'vacuum': {'before_bytes': Value('int64'), 'after_bytes': Value('int64'), 'seconds': Value('float64')}}
              because column names don't match

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.

Pocketpedia prebuilt indexes

Ready-to-use offline search indexes of English Wikipedia for Pocketpedia, so small local language models can look things up without the 11.6 GB download and build step. Each tier holds all 6,407,814 articles of the wikimedia/wikipedia 20231101.en snapshot as SQLite (FTS5 search index plus zstd-compressed article text).

Tier	Size	PopQA (held-out 1,000)	SimpleQA correct (300)	Notes
`lean` (default)	8.5 GiB	52.3%	39.0%	trained zstd dictionary, reference sections stripped, articles capped at 4,000 words
`phone`	5.8 GiB	50.7%	33.3%	as lean, articles capped at 500 words
`lite`	4.5 GiB	38.3%	14.3%	titles and intros only, keyword retrieval (v1)
full (build it yourself)	14.4 GiB	52.8%	40.3%

Scores are Gemma 3 4B answering with Pocketpedia's retrieval, on the pre-registered benchmark. From memory alone the same model scores 16.0% on PopQA. SimpleQA is graded by a local GPT-OSS 120B with the official template, so it is not comparable with published SimpleQA scores.

Use

These indexes need Pocketpedia 0.2 or later (the lean and phone tiers use build options added in 0.2).

pocketpedia setup --tier lean       # downloads and verifies the tier into ~/.pocketpedia

Or download by hand, keeping the tier folder as the index folder:

hf download Tetsuto/pocketpedia-index --repo-type dataset --include "lean/*" --local-dir ./pocketpedia-data

Each tier folder has a manifest.json with file sizes, SHA-256 checksums and the build options; the root manifest.json lists the tiers.

Licence and attribution

The text is from Wikipedia, written by its contributors, and is licensed under the Creative Commons Attribution-ShareAlike 4.0 License (some text is also available under the GFDL; see Wikipedia's copyright notice). These indexes are an adaptation: the text was split, trimmed (reference sections removed, long articles truncated in the lean and phone tiers) and compressed. They are shared under the same licence. When you quote or redistribute text from them, credit Wikipedia and its contributors, link the source article, indicate changes, and keep the share-alike terms. The live articles may differ from this 2023 snapshot.

Built from wikimedia/wikipedia (20231101.en) with pocketpedia build --profile lean|phone|lite. Pocketpedia's code is MIT licensed.

Downloads last month: 108

Total file size:

47.9 GB