XiaoHongbo-Hope opened a new pull request, #10758: URL: https://github.com/apache/paimon/pull/10758
### Purpose LeRobot state/action tensors currently map to `VECTOR`, which places them in separate vector files under data evolution. These training features should use `ARRAY` so they remain with ordinary row data and can use ROW sidecars. Map new imports to `ARRAY` (nested for higher-rank tensors), preserving shape metadata and value validation. Existing `VECTOR` datasets remain readable for training. The general Arrow-to-Paimon mapping is unchanged; existing tables are not migrated. No performance claim is made in this PR. ### Tests - LeRobot 0.4.4: 98 passed, 1 skipped, covering import schemas/values, video file layout, training reads and legacy VECTOR tables. - LeRobot 0.6.1 validation suite: 69 passed, 1 skipped, 78 subtests passed. Full import tests remain blocked by the DatasetInfo compatibility failures also reproduced on master; the added legacy-import test hits the same limitation. - Flake8, Python 3.6 syntax and `git diff --check` passed. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
