GloVe icon indicating copy to clipboard operation
GloVe copied to clipboard

URL with GloVe dataset not working

Open Nayef211 opened this issue 4 years ago • 6 comments

  • URL for GloVe.840B.300d not working https://nlp.stanford.edu/data/glove.840B.300d.zip

Nayef211 avatar Jul 08 '20 14:07 Nayef211

Our cluster is currently down for maintenance. It should be back within a day.

AngledLuffa avatar Jul 08 '20 17:07 AngledLuffa

We have disabled our tests again due to the URL issue. https://github.com/pytorch/text/pull/888

zhangguanheng66 avatar Jul 17 '20 15:07 zhangguanheng66

I'm a little confused by that comment, because aren't those tests downloading mock zip files instead of the original zip files on nlp.stanford.edu?

Regardless, the nlp.stanford.edu url is working again.

AngledLuffa avatar Jul 17 '20 16:07 AngledLuffa

Disabled tests again due to URL breakage https://github.com/pytorch/text/pull/920.

To answer your question @AngledLuffa, we are using mock zip files for our tests, however, the filename is found by getting the headers from a HTTP GET call. You can refer to this code snippet for more info.

Nayef211 avatar Aug 12 '20 17:08 Nayef211

Can I recommend mocking that as well, then? Because our hardware occasionally has problems, and while we obviously want to fix it as quickly as possible, we also don't want to be responsible for your unit tests not working.

AngledLuffa avatar Aug 12 '20 17:08 AngledLuffa

Can I recommend mocking that as well, then? Because our hardware occasionally has problems, and while we obviously want to fix it as quickly as possible, we also don't want to be responsible for your unit tests not working.

Yeap. That's what we are planning to do. But we probably won't moch the entire datasets because we don't want to host the datasets for legal issue.

zhangguanheng66 avatar Aug 12 '20 17:08 zhangguanheng66