Brain training data extraction issues


I want to train models using the brain training data. I downloaded it using curl with recovery mode turned on (as suggested in the data links mail).
The download is complete (re-executing the curl command doesn’t have any effect).

However, when extracting the archive using tar -xvf brain_multicoil_train.tar.gz -C $SCRATCH/brain_multicoil_train/, I encounter the following error:

tar: Unexpected EOF in archive
tar: Unexpected EOF in archive
tar: Error is not recoverable: exiting now

I am right now in the process of checking the sha256 checksum, but since it’s an extremely large file, it takes some time… I will update the topic with the info once I have it, but I already wanted to know if anyone else had had trouble with the brain training data (especially since I see that only the fastMRI team posted a submission for the brain leaderboard).


Just got the checksum result, and it’s matching what’s in the SHA256 file I got from the data links mail. So I guess there is a problem with the archive itself. (Just a small note, it took 40min to get the checksum !!)


Actually, this was just a problem of my process being killed since I was on the front node of an HPC cluster and the extraction was taking a long time. I didn’t realize it because of the peculiar error message, my bad.
It now works fine.

Great to hear - thanks for the information, Zaccharie.