Torch.hub.load() gives "HTTPError: rate limit exceeded"

D_L · June 22, 2021, 9:42pm

Working through the deep insight example herehttps://github.com/alok-ai-lab/DeepInsight/blob/master/examples/pytorch_squeezenet.ipynb.

However, when trying to load this I get “HTTPError: rate limit exceeded”

I did some searching and couldn’t find a solution. There were some mentions of something similar being due to your location here.

Any thoughts?

D_L · June 23, 2021, 7:21pm

FYI - work around is downloading to disk from git and:

torch.hub.load(PATH , ‘squeezenet1_1’, source = ‘local’, pretrained=False, verbose=False)

cloudhan · June 24, 2021, 3:51am

You can comment this line out to workaround this issue temporarily.

github.com

pytorch/pytorch/blob/80f40b172f4d1491713d6608f2196e59eb04b1b5/torch/hub.py#L162

    
      
          # To check if cached repo exists, we need to normalize folder names.
          repo_dir = os.path.join(hub_dir, '_'.join([repo_owner, repo_name, normalized_br]))
          
          
use_cache = (not force_reload) and os.path.exists(repo_dir)
          
          
if use_cache:
              if verbose:
                  sys.stderr.write('Using cache found in {}\n'.format(repo_dir))
          else:
              # Validate the tag/branch is from the original repo instead of a forked repo
              _validate_not_a_forked_repo(repo_owner, repo_name, branch)
          
          
    cached_file = os.path.join(hub_dir, normalized_br + '.zip')
              _remove_if_exists(cached_file)
          
          
    url = _git_archive_link(repo_owner, repo_name, branch)
              sys.stderr.write('Downloading: \"{}\" to {}\n'.format(url, cached_file))
              download_url_to_file(url, cached_file, progress=False)
          
          
    with zipfile.ZipFile(cached_file) as cached_zipfile:
                  extraced_repo_name = cached_zipfile.infolist()[0].filename

Verner · July 6, 2021, 8:09pm

Hi D_L,

Did you get a solution to this problem?

D_L · July 6, 2021, 10:13pm

I worked around using the solution I posted above. Thanks for following up!

Verner · July 6, 2021, 11:28pm

Hi again,

Thanks for answering. Did you download the code and use it stored on your local machine? That’s it?
Thanks a lot.

Verner · July 7, 2021, 11:55pm

Hi D_L,
Thanks for answer.
I managed to run the model by copying the code from github to Collab. I did this after reading your answer.
It is not a definitive solution, but it serves to continue my research.
Thanks a lot.

durgeshrathod777 · June 25, 2024, 12:16pm

Pytorch released a bugfix for 1.9 (i.e. 1.9.1) so updating to 1.9.1 should rectify this issue for good