Broken links for (transformer-based) translation models with torch.hub.load

🐛 Bug

Trying to load various transformer-based models for translations as documented in several READMEs (see below) with torch.hub.load fails.

For PyTorch version 1.8.1 (from pip): The step Downloading: "https://github.com/pytorch/fairseq/archive/master.zip" fails with an HTTP Error 404: urllib.error.HTTPError: HTTP Error 404: Not Found

Using PyTorch version 1.9.1 (from pip) instead yields the following error message: ValueError: Cannot find master in https://github.com/pytorch/fairseq. If it's a commit from a forked repo, please call hub.load() with forked repo directly.

To Reproduce

The following snippets are all copy&pasted from READMEs.

import torch
en2fr = torch.hub.load('pytorch/fairseq', 'transformer.wmt14.en-fr', tokenizer='moses', bpe='subword_nmt')

import torch
en2de = torch.hub.load('pytorch/fairseq', 'transformer.wmt19.en-de.single_model')

import torch
print(torch.hub.list('pytorch/fairseq'))

Code sample

import torch
en2de = torch.hub.load('pytorch/fairseq', 'transformer.wmt19.en-de.single_model')

Expected behavior

Models should be downloaded to local cache so that translation can be started.

Environment

fairseq Version (e.g., 1.0 or main): 0.10.2
PyTorch Version (e.g., 1.0): 1.8.1 / 1.9.1
OS (e.g., Linux): MacOS (11.6)
How you installed fairseq (pip, source): pip
Build command you used (if compiling from source): N/A
Python version: 3.7.10
CUDA/cuDNN version: N/A
GPU models and configuration: N/A
Any other relevant information: N/A

Additional context

N/A

facebookresearch / fairseq