Datasets:

---

AI4Spectro

/

ILtrainV1_P2

like 0

Follow

SimXRD 8

I am writing to report a consistent technical issue specifically with the ILtrainV1_P2 dataset. Our research group is currently utilizing the IL dataset series, and we have confirmed that other segments, such as ILtrainV1_P1, work perfectly under our current workflow. The issue is localized exclusively to the ILtrainV1_P2 repository.

After downloading the five split parts of ILtrainV1_P2 and performing the merge/decompression as specified in the README, the resulting database file is unreadable. Specifically, the system returns a "database disk image is malformed" error. Detailed logs indicate a B-tree initialization failure, such as: "Tree 7 page 37912941: btreeInitPage() returns error code 11".

To ensure this was not a local transmission issue, we conducted a rigorous troubleshooting process. We re-downloaded all five components of P2 multiple times across different devices and network environments. Most importantly, we have verified the SHA256 checksums for each of the five parts, and they match the original hash values hosted on Hugging Face perfectly. This confirms that the files were downloaded without corruption, yet the combined database remains functionally broken.

Given that the rest of the IL dataset series functions correctly, it appears the issue resides in the source files of P2 or the specific archiving process used for this part. As this dataset is essential for our research in self-supervised pre-training, we would greatly appreciate it if you could verify the integrity of the uploaded P2 parts or investigate the internal database structure. We are available to provide more extensive error logs if it would assist in resolving this matter.