Skip to content

Pin the khm-latn corpus provenance (WO68 lineage recovery) - #388

Merged
ronaldtse merged 1 commit into
mainfrom
feat/khm-latn-provenance
Oct 29, 2026
Merged

ronaldtse merged 1 commit into
mainfrom
feat/khm-latn-provenance

Conversation

@ronaldtse

Copy link
Copy Markdown
Contributor

Summary

Resolves the UNVERIFIED blockers in corpora/khm-latn.yaml: the training corpus is identified exactly as secryst/data-khmer-translit (17,911 word pairs, two parallel CSVs), staged to the training bucket with SHA-256 checksums, retiring the legacy raw.githubusercontent fetch. The upstream repo has no license file (predecessor project's own), so the ship gate remains owner-held until a license_review record lands.

Recovery run run-081-khm-latn-recovery launched (seed 42, the legacy recipe: ByT5-small, 30 epochs, early stopping) against khm-latn-1.0's 27.4 CER.

…t identified and staged

The WO68 lineage recovery: the training corpus is exactly
secryst/data-khmer-translit (17,911 pairs), staged to the training
bucket with sha256, retiring the raw.githubusercontent fetch. The
upstream repo has no license file, so the ship gate stays owner-held.
Recovery run-081 launched (seed 42, legacy recipe).
@ronaldtse
ronaldtse merged commit 8072d19 into main Oct 29, 2026
7 of 11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant