Battam's picture
Release release-2026-11-17.3, made by the table store from release-2026-11-17.2 (part 5)
3f52794 verified
|
Raw History Blame Contribute Delete
4.01 kB

ModelScope datasets and models (modelscope)

Kind: corpus (raw). One whole upstream that a collector keeps reading in bulk, counted in rows: each row points to one file or record.

Flagged, not usable as it is: no collection header yet: title, description, licence and fetch fields are empty.

Files in public dataset and model repositories on ModelScope (modelscope.cn), listed through the public API used by the official ModelScope SDK. One pointer per file; uri is the official download API pinned to the repository commit, and size and sha256 come from the repository file tree. Type, namespace, name, version, license field, task tags, downloads, last update time and path are in source_meta.

Summary from sample files: not judged yet. The text above is the source's own description.

Domain and form

  • Domain: D4.5 (paper reproduction and classic ML), D4.4 (evaluation and monitoring).
  • Form: model-hubs. Model and dataset hubs.

Unit and scale

  • One row points at: file.
  • Rows: 8,726,153.
  • Size: Recorded size 2.03 PB; deduplicated size 1.86 PB; size check: 100.0% match (299 of the 300 rows read compared); not compared: file gone 1.
  • Fetch check: 299 of the 300 sampled rows read (99.7%) fetched and matched their recorded size and hash (read 2026-11-13); the others did not.
  • Batches: 1 (all). Collection: still being collected.

File types and sizes

Type Rows Size
mp4 1,430,765 44.4 TB
jpg 1,376,639 652 GB
parquet 1,197,813 468 TB
png 923,340 969 GB
json 502,739 1.31 TB
wav 340,175 807 GB
hdf5 271,784 85.7 TB
tar 257,882 610 TB
txt 232,379 86.3 GB
zip 187,110 175 TB
other types: 191

Size of one row, upper bounds by quantile: 10% 4.09 KB, 50% 524 KB, 90% 537 MB, 99% 8.59 GB.

License and access

  • License: Per repository (see license_override); license tier mixed; training use: yes 85%, unknown 10%, conditional 5%; redistribution: yes 83%, unknown 10%, conditional 7%, no <1%; files may be shipped: mixed.
  • License statement: not recorded on the card.
  • Login: no.
  • How to fetch: not recorded on the card; the row's uri is fetched as it is.

Freshness

  • Rows are counted for the release of 2026-11-17; the card's own sample and counts date from 2026-10-08.

Readiness

  • Readiness level: not judged yet.
  • Card review: draft (a draft card has not been reviewed as a whole).
  • Personal information read from samples: not judged yet.

Query this source at a fixed release

Pinned to the release tag release-2026-11-17.3.

SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-17.3/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-17.3/meta/catalog.parquet' WHERE source = 'modelscope'));
SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
FROM read_parquet(getvariable('files'))
WHERE source = 'modelscope'
LIMIT 10;

Sample files

No sample files are recorded on the card yet.

Parts not to use for tasks

  • Parts to leave out: not judged yet.
  • Benchmark overlap task_file (scope row; benchmarks: DeepSWE, FrontierSWE v2, GDPval, GraphWalks, LHTB, LibriSpeech, MCP Atlas, OSWorld, OSWorld 2, SWE-bench Multilingual, SWE-bench Pro, SWE-bench Verified, SciCode, Terminal-Bench, Toolathlon, WideSearch, agentslastexam, apexagents, browsecomp, deepswe, gdpval, mcpatlas, multiswe, osworld, scicode, swebench, tau2, tau2-bench, terminalbench, widesearch): the default profile of tools/fetch.py removes 52 rows and 43 whole collections.

The card modelscope.yaml (card_version 5) holds every field with its evidence; values shown here as not judged yet have no judgment on the card so far.