Datasets:

---

PipeMFL

/

PipeMFL-240K

Like 3

Follow

PipeMFL 4

Hello!

I'm trying to download your PipeMFL-240K dataset. And all large files (dataset train, test & val parts) download with progress showing 100% but all file saved in my storage are missing ~1-2GB at the end. This happens with every file - while extracting files through the CLI, the system freezes, and WinRAR extracts the most part of the data, but at the end it returns an error that the archive is corrupted. As a result, about 10% of the images and all the markup are missing.

What I tried:

• hf_hub_download() with resume_download=True (python script)
• snapshot_download() with max_workers=1 (python script)
• huggingface-cli download
• git clone + git lfs pull
• Removing cache before each attempt
  Result always the same.

My environment:

• Windows 11
• Python 3.9
• hf 1.8.0 (huggingface_hub)

Could you provide:

1. Direct HTTP links for the files (bypassing LFS/Xet)?
2. Or a torrent/magnet link?
3. Or any known working method for Windows users?

If the files themselves on the server are incomplete, please let me know.

Additional info: This doesn't seem to be the known Xet race condition - files aren't corrupted with random data, they're simply cut off at the end. The last chunk appears to be missing entirely.

Thank you for your time!