plot_version_growth: v0.2.5 totals from the tagged CHANGELOG (20 sources)

#89
by ppuzio - opened
Fabryka AI org

src/plot_version_growth.py plotted v0.2.5 at the 18-source draft totals: 4,319,200 documents / 9,639,062,573 tokens. The CHANGELOG at the v0.2.5 tag (lines 32-35) gives 4,366,651 / 9,676,513,980 across 20 sources. The draft totals were from before sejm_api and wiktionary_examples were counted.

verify_paper_claims.py (PR #88) uses these literals as the growth series. With the old literals it reports 8 FAILs at the paper's \snapshot pin. All of them come from this point: the v0.2.5 documents and tokens rows, the v0.2.5 additions sums and Total added, and the growth anchor.

Changes:

  • src/plot_version_growth.py: v0.2.5 DOCUMENTS and TOKENS now match the tagged CHANGELOG.
  • artifacts/dataset_size_by_version.png: regenerated (LFS). README.md embeds it, and README.md itself is unchanged.

No data changes, so there is no render_sample.py output. pytest from the repo root: 616 passed.

The pin keeps failing after this merges, because the v0.2.5 tag can't be changed. On main, --ref worktree still fails on the v0.2.6 changes until the paper is updated after the dataset freeze.

🤖 Generated with Claude Code

ppuzio changed pull request status to open
ppuzio changed pull request status to merged
ppuzio deleted the refs/pr/89 ref

Sign up or log in to comment