XiaoHongbo-Hope opened a new pull request, #10821: URL: https://github.com/apache/paimon/pull/10821
### Purpose Explicit TorchCodec reads currently use exact mode and scan the video during initialization, even when a persisted keyframe index is available. Use the MP4 sample table to map frame ordinals to timestamps, prefetch metadata and target GOP ranges, and verify the returned PTS before returning frames. Unsupported layouts or failed verification fall back to exact mode. Parsing and prefetching are bounded. Applies to `video_backend="torchcodec"`; the default backend is unchanged. Indexed TorchCodec also respects `read.video.codec-threads`. ### Tests Real TorchCodec coverage for B-frames, VFR, audio-first and multi-video MP4s, fragmented-video fallback, incorrect PTS, and parser bounds; existing indexed-reader regressions. Full DataLoader validation on local storage and OSSFS compares sample order and first/last batch tensors with Lance. -- 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]
