def main():
mp.freeze_support()
args = parser.parse_args()
fix_path()
if args.subcommand == 'align':
run_align_corpus(args, acoustic_languages)
elif args.subcommand == 'train':
run_train_corpus(args)
elif args.subcommand == 'g2p':
run_g2p(args, g2p_languages)
elif args.subcommand == 'train_g2p':
run_train_g2p(args)
elif args.subcommand == 'validate':
run_validate_corpus(args)
elif args.subcommand == 'download':
run_download(args)
elif args.subcommand == 'train_lm':
run_train_lm(args)
elif args.subcommand == 'train_ivector':
run_train_ivector_extractor(args)
elif args.subcommand == 'annotator':
run_annotator(args)
elif args.subcommand == 'thirdparty':
run_thirdparty(args)
elif args.subcommand == 'transcribe':
run_transcribe_corpus(args)
unfix_path()
def test_transcribe(
basic_corpus_dir,
basic_dict_path,
english_acoustic_model,
generated_dir,
transcription_acoustic_model,
transcription_language_model,
temp_dir,
transcribe_config_path,
):
output_path = os.path.join(generated_dir, "transcribe_test")
command = [
"transcribe",
basic_corpus_dir,
basic_dict_path,
transcription_acoustic_model,
transcription_language_model,
output_path,
"-t",
temp_dir,
"-q",
"--clean",
"--debug",
"-v",
"--config_path",
transcribe_config_path,
]
args, unknown = parser.parse_known_args(command)
run_transcribe_corpus(args)
assert os.path.exists(os.path.join(output_path, "michael", "acoustic_corpus.lab"))
def test_transcribe_speaker_dictionaries(
multilingual_ipa_corpus_dir,
mfa_speaker_dict_path,
english_mfa_acoustic_model,
generated_dir,
transcription_language_model,
temp_dir,
transcribe_config_path,
):
output_path = os.path.join(generated_dir, "transcribe_test")
command = [
"transcribe",
multilingual_ipa_corpus_dir,
mfa_speaker_dict_path,
english_mfa_acoustic_model,
transcription_language_model,
output_path,
"-t",
temp_dir,
"-q",
"--clean",
"--debug",
"--config_path",
transcribe_config_path,
]
args, unknown = parser.parse_known_args(command)
run_transcribe_corpus(args, unknown)
def test_transcribe_arpa(
basic_corpus_dir,
basic_dict_path,
english_acoustic_model,
generated_dir,
transcription_language_model_arpa,
temp_dir,
transcribe_config_path,
):
if sys.platform == "win32":
pytest.skip("No LM generation on Windows")
temp_dir = os.path.join(temp_dir, "arpa_test_temp")
output_path = os.path.join(generated_dir, "transcribe_test_arpa")
print(transcription_language_model_arpa)
command = [
"transcribe",
basic_corpus_dir,
basic_dict_path,
english_acoustic_model,
transcription_language_model_arpa,
output_path,
"-t",
temp_dir,
"-q",
"--clean",
"--debug",
"-v",
"--use_mp",
"false",
"--config_path",
transcribe_config_path,
]
args, unknown = parser.parse_known_args(command)
run_transcribe_corpus(args)
assert os.path.exists(os.path.join(output_path, "michael", "acoustic_corpus.lab"))
def test_transcribe(basic_corpus_dir, sick_dict_path, english_acoustic_model,
generated_dir, transcription_acoustic_model,
transcription_language_model, temp_dir, transcribe_config):
output_path = os.path.join(generated_dir, 'transcribe_test')
command = [
'transcribe', basic_corpus_dir, sick_dict_path,
transcription_acoustic_model, transcription_language_model,
output_path, '-t', temp_dir, '-q', '--clean', '-d', '--config',
transcribe_config
]
args, unknown = parser.parse_known_args(command)
run_transcribe_corpus(args)
def test_transcribe(basic_corpus_dir, sick_dict_path, english_acoustic_model,
generated_dir, transcription_acoustic_model,
transcription_language_model, temp_dir):
output_path = os.path.join(generated_dir, 'transcribe_test')
args = DummyArgs()
args.acoustic_model_path = transcription_acoustic_model
args.corpus_directory = basic_corpus_dir
args.dictionary_path = sick_dict_path
args.language_model_path = transcription_language_model
args.output_directory = output_path
args.temp_directory = temp_dir
args.evaluate = True
run_transcribe_corpus(args)
def main() -> None:
"""
Main function for the MFA command line interface
"""
check_third_party()
hooks = ExitHooks()
hooks.hook()
atexit.register(hooks.history_save_handler)
from colorama import init
init()
parser = create_parser()
mp.freeze_support()
args, unknown = parser.parse_known_args()
for short in ["-c", "-d"]:
if short in unknown:
print(
f"Due to the number of options that `{short}` could refer to, it is not accepted. "
"Please specify the full argument",
file=sys.stderr,
)
sys.exit(1)
try:
if args.subcommand in ["g2p", "train_g2p"]:
try:
import pynini # noqa
except ImportError:
print(
"There was an issue importing Pynini, please ensure that it is installed. If you are on Windows, "
"please use the Windows Subsystem for Linux to use g2p functionality.",
file=sys.stderr,
)
sys.exit(1)
if args.subcommand == "align":
run_align_corpus(args, unknown)
elif args.subcommand == "adapt":
run_adapt_model(args, unknown)
elif args.subcommand == "train":
run_train_acoustic_model(args, unknown)
elif args.subcommand == "g2p":
run_g2p(args, unknown)
elif args.subcommand == "train_g2p":
run_train_g2p(args, unknown)
elif args.subcommand == "validate":
run_validate_corpus(args, unknown)
elif args.subcommand == "validate_dictionary":
run_validate_dictionary(args, unknown)
elif args.subcommand in ["model", "models"]:
run_model(args)
elif args.subcommand == "train_lm":
run_train_lm(args, unknown)
elif args.subcommand == "train_dictionary":
run_train_dictionary(args, unknown)
elif args.subcommand == "train_ivector":
run_train_ivector_extractor(args, unknown)
elif args.subcommand == "classify_speakers": # pragma: no cover
run_classify_speakers(args, unknown)
elif args.subcommand in ["annotator", "anchor"]:
run_anchor()
elif args.subcommand == "transcribe":
run_transcribe_corpus(args, unknown)
elif args.subcommand == "create_segments":
run_create_segments(args, unknown)
elif args.subcommand == "configure":
update_global_config(args)
global GLOBAL_CONFIG
GLOBAL_CONFIG = load_global_config()
elif args.subcommand == "history":
print_history(args)
elif args.subcommand == "version":
from montreal_forced_aligner.utils import get_mfa_version
print(get_mfa_version())
elif args.subcommand == "thirdparty": # Deprecated command
raise DeprecationWarning(
"Necessary thirdparty executables are now installed via conda. Please refer to the installation docs for the updated commands."
)
elif args.subcommand == "download": # Deprecated command
raise DeprecationWarning(
"Downloading models is now run through the `mfa model download` command, please use that instead."
)
except MFAError as e:
if getattr(args, "debug", False):
raise
print(e, file=sys.stderr)
sys.exit(1)
def main():
mp.freeze_support()
args, unknown = parser.parse_known_args()
fix_path()
if args.subcommand in ['align', 'train', 'train_ivector']:
from montreal_forced_aligner.thirdparty.kaldi import validate_alignment_binaries
if not validate_alignment_binaries():
print(
"There was an issue validating Kaldi binaries, please ensure you've downloaded them via the "
"'mfa thirdparty download' command. See 'mfa thirdparty validate' for more detailed information "
"on why this check failed.")
sys.exit(1)
elif args.subcommand in ['transcribe']:
from montreal_forced_aligner.thirdparty.kaldi import validate_transcribe_binaries
if not validate_transcribe_binaries():
print(
"There was an issue validating Kaldi binaries, please ensure you've downloaded them via the "
"'mfa thirdparty download' command. See 'mfa thirdparty validate' for more detailed information "
"on why this check failed. If you are on MacOS, please note that the thirdparty binaries available "
"via the download command do not contain the transcription ones. To get this functionality working "
"for the time being, please build kaldi locally and follow the instructions for running the "
"'mfa thirdparty kaldi' command.")
sys.exit(1)
elif args.subcommand in ['train_dictionary']:
from montreal_forced_aligner.thirdparty.kaldi import validate_train_dictionary_binaries
if not validate_train_dictionary_binaries():
print(
"There was an issue validating Kaldi binaries, please ensure you've downloaded them via the "
"'mfa thirdparty download' command. See 'mfa thirdparty validate' for more detailed information "
"on why this check failed. If you are on MacOS, please note that the thirdparty binaries available "
"via the download command do not contain the train_dictionary ones. To get this functionality working "
"for the time being, please build kaldi locally and follow the instructions for running the "
"'mfa thirdparty kaldi' command.")
sys.exit(1)
elif args.subcommand in ['g2p', 'train_g2p']:
try:
import pynini
except ImportError:
print(
"There was an issue importing Pynini, please ensure that it is installed. If you are on Windows, "
"please use the Windows Subsystem for Linux to use g2p functionality."
)
sys.exit(1)
if args.subcommand == 'align':
run_align_corpus(args, unknown, acoustic_languages)
elif args.subcommand == 'train':
run_train_corpus(args)
elif args.subcommand == 'g2p':
run_g2p(args, g2p_languages)
elif args.subcommand == 'train_g2p':
run_train_g2p(args)
elif args.subcommand == 'validate':
run_validate_corpus(args)
elif args.subcommand == 'download':
run_download(args)
elif args.subcommand == 'train_lm':
run_train_lm(args)
elif args.subcommand == 'train_dictionary':
run_train_dictionary(args)
elif args.subcommand == 'train_ivector':
run_train_ivector_extractor(args)
elif args.subcommand == 'classify_speakers':
run_classify_speakers(args)
elif args.subcommand == 'annotator':
from montreal_forced_aligner.command_line.annotator import run_annotator
run_annotator(args)
elif args.subcommand == 'thirdparty':
run_thirdparty(args)
elif args.subcommand == 'transcribe':
run_transcribe_corpus(args)
elif args.subcommand == 'create_segments':
run_create_segments(args, unknown)
elif args.subcommand == 'version':
print(__version__)
unfix_path()
