Originally these labels were available at ImageNet website. The website now returns invalid page. Moreover, downloading the original data from ImageNet website is painfully slow. Downloading the validation set from AcademicTorrents is fast enough for everyone's need.
The only pain is that you don't get original labels from ImageNet website.
Fortunately, we can solve this using following piece of python code: