Cannot load XLM from torch hub - cannot unpack non-iterable NoneType object

[arpcode]

xlmr = torch.hub.load('pytorch/fairseq', 'xlmr.large')

I get the following error message -

TypeError                                 Traceback (most recent call last)
/tmp/ipykernel_11808/3675591151.py in <module>
----> 1 xlmr = torch.hub.load('pytorch/fairseq', 'xlmr.large', force_reload = True)

~/anaconda3/envs/btp/lib/python3.9/site-packages/torch/hub.py in load(repo_or_dir, model, source, force_reload, verbose, skip_validation, *args, **kwargs)
    397         repo_or_dir = _get_cache_or_reload(repo_or_dir, force_reload, verbose, skip_validation)
    398 
--> 399     model = _load_local(repo_or_dir, model, *args, **kwargs)
    400     return model
    401 

~/anaconda3/envs/btp/lib/python3.9/site-packages/torch/hub.py in _load_local(hubconf_dir, model, *args, **kwargs)
    423 
    424     hubconf_path = os.path.join(hubconf_dir, MODULE_HUBCONF)
--> 425     hub_module = import_module(MODULE_HUBCONF, hubconf_path)
    426 
    427     entry = _load_entry_from_hubconf(hub_module, model)

~/anaconda3/envs/btp/lib/python3.9/site-packages/torch/hub.py in import_module(name, path)
     74     module = importlib.util.module_from_spec(spec)
     75     assert isinstance(spec.loader, Loader)
---> 76     spec.loader.exec_module(module)
     77     return module
     78 

~/anaconda3/envs/btp/lib/python3.9/importlib/_bootstrap_external.py in exec_module(self, module)

~/anaconda3/envs/btp/lib/python3.9/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

~/.cache/torch/hub/pytorch_fairseq_main/hubconf.py in <module>
     37 
     38 # only do fairseq imports after checking for dependencies
---> 39 from fairseq.hub_utils import (  # noqa; noqa
     40     BPEHubInterface as bpe,
     41     TokenizerHubInterface as tokenizer,

~/.cache/torch/hub/pytorch_fairseq_main/fairseq/__init__.py in <module>
     31 hydra_init()
     32 
---> 33 import fairseq.criterions  # noqa
     34 import fairseq.distributed  # noqa
     35 import fairseq.models  # noqa

~/.cache/torch/hub/pytorch_fairseq_main/fairseq/criterions/__init__.py in <module>
     16 
     17 
---> 18 (
     19     build_criterion_,
     20     register_criterion,

TypeError: cannot unpack non-iterable NoneType object

I've seen this posted time and again, but none of the threads provide a solution.

[YanineeJam]

You should install the suitable torch version for this command. I recommend installing torch==1.6.0 torchvision==0.7.0. It works for me. :)

[zhouliang-yu]

I have the same problem. I tried to upgrade the torch and torchvision version to 1.6.0 and 0.7.0 respectively but it doesn't help

[shoang22]

Did you try installing via pip install fairseq? I ran into the same issue when I tried to pip install the editable version. I had to uninstall it and reinstall with the aforementioned command.

[VidushiVashishth]

pip installing fairseq changes the error to the following for me:

[/usr/lib/python3.8/distutils/dist.py](https://localhost:8080/#) in _parse_command_opts(self, parser, args)
    544         # to be sure that the basic "command" interface is implemented.
    545         if not issubclass(cmd_class, Command):
--> 546             raise DistutilsClassError(
    547                 "command class %s must subclass Command" % cmd_class)
    548 

DistutilsClassError: command class <class 'torch.utils.cpp_extension.BuildExtension'> must subclass Command