There are some very tricky implementations in the
Fairseq. And I am confused about the [models, criterions, tasks]-specific parser used in
__init__.py scripts, eg: models/__init__.py. There are some parsers, like
lstm_parser, are declared, but they seem impossible to actually used in command line for the help information and argument parsing.
Is there any way to show the [models, tasks, criterions]-specific arguments in command line?