SimpleLSTM faiseq-train tutorial cannot import 'metrics'

[Efrainq07]

🐛 Bug

Following the SimpleLSTM tutorial, everything works fine until fairseq-train is used as follows:

fairseq-train data-bin/iwslt14.tokenized.de-en \ --arch tutorial_simple_lstm \ --encoder-dropout 0.2 --decoder-dropout 0.2 \ --optimizer adam --lr 0.005 --lr-shrink 0.5 \ --max-tokens 12000

Which gives the following error message:

  File "<string>", line 1, in <module>
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/multiprocessing/spawn.py", line 116, in spawn_main
    exitcode = _main(fd, parent_sentinel)
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/multiprocessing/spawn.py", line 125, in _main
    prepare(preparation_data)
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/multiprocessing/spawn.py", line 236, in prepare
    _fixup_main_from_path(data['init_main_from_path'])
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/multiprocessing/spawn.py", line 287, in _fixup_main_from_path
    main_content = runpy.run_path(main_path,
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/runpy.py", line 288, in run_path
    return _run_module_code(code, init_globals, run_name,
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/runpy.py", line 97, in _run_module_code
    _run_code(code, mod_globals, init_globals,
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/Users/efrainq/opt/anaconda3/bin/fairseq-train", line 5, in <module>
    from fairseq_cli.train import cli_main
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq_cli/train.py", line 40, in <module>
    from fairseq.model_parallel.megatron_trainer import MegatronTrainer
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/model_parallel/__init__.py", line 6, in <module>
    from . import criterions, models, modules  # noqa
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/model_parallel/criterions/__init__.py", line 14, in <module>
    importlib.import_module("fairseq.model_parallel.criterions." + module)
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/model_parallel/criterions/vocab_parallel_cross_entropy.py", line 10, in <module>
    from fairseq.criterions import FairseqCriterion, register_criterion
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/criterions/__init__.py", line 36, in <module>
    importlib.import_module("fairseq.criterions." + file_name)
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/criterions/speech_dlm_criterion.py", line 11, in <module>
    from fairseq import metrics, utils
ImportError: cannot import name 'metrics' from 'fairseq' (unknown location)
^CTraceback (most recent call last):
  File "/Users/efrainq/opt/anaconda3/bin/fairseq-train", line 8, in <module>
    sys.exit(cli_main())
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq_cli/train.py", line 574, in cli_main
    distributed_utils.call_main(cfg, main)
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/distributed/utils.py", line 404, in call_main
    main(cfg, **kwargs)
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq_cli/train.py", line 189, in main
    for _ in itr:
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/data/iterators.py", line 57, in __next__
    x = next(self._itr)
  File "/Users/efrainq/Documents/Repositories/fairseq/fairseq/data/iterators.py", line 745, in __next__
    item = self._queue.get(True)
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/queue.py", line 171, in get
    self.not_empty.wait()
  File "/Users/efrainq/opt/anaconda3/lib/python3.9/threading.py", line 312, in wait
    waiter.acquire()

To Reproduce

1. Follow the tutorial and create the LSTM example files
2. Call fairseq-train using

fairseq-train data-bin/iwslt14.tokenized.de-en \
  --arch tutorial_simple_lstm \
  --encoder-dropout 0.2 --decoder-dropout 0.2 \
  --optimizer adam --lr 0.005 --lr-shrink 0.5 \
  --max-tokens 12000

4. See the traceback error related to a missing metrics import.

Expected behavior

Training should run seamlessly for the example in the tutorial.

Additional context

This can be fixed just by changing the metrics import in fairseq/fairseq/criterions/speech_dlm_criterion.py to the new fairseq.logging location.