'speaker_name_width'))
max_speaker_name_width = max(corp_speaker_name_width,
speaker_name_width)
corpora.append([
config.get("corpus" + str(cur_corpus_id), "location"),
config.get("corpus" + str(cur_corpus_id), "prefix"),
config.get("corpus" + str(cur_corpus_id), "word_suffix"),
corp_speaker_name_width
])
cur_corpus_id += 1
if len(corpora) == 0: sys.exit("No Corpora Found!")
data_manipulation.import_dictionaries(dicts)
data_manipulation.import_corpora(corpora, max_speaker_name_width)
mkmono0_led = 'config/mkmono0.led'
mkmono1_led = 'config/mkmono1.led'
# Make phone transcriptions
if not os.path.exists(mkmono0_led): sys.exit("Not Found: " + mkmono0_led)
if not os.path.exists(mkmono1_led): sys.exit("Not Found: " + mkmono1_led)
if os.path.isdir('files'): shutil.rmtree('files')
os.mkdir('files')
excluded = []
if config.getint('corpora', 'skip_dict_lookup') == 1:
shutil.copyfile('corpora/words.mlf', 'files/words.mlf')
else:
excluded = data_manipulation.prune_transcriptions(
parser.add_option('-q', '--queue', dest="queue", help="Fix the queue where things are running in", default="")
options, configs = parser.parse_args()
if len(options.recognition_dir) == 0:
dirs_done = []
if len(options.dirs_done) > 0:
dirs_done = options.dirs_done.split(',')
experiments, model = parse_config(configs)
model.make_reference()
if model.configuration['dict'] is not None:
data_manipulation.import_dictionaries([(model.configuration['dict'], '', '')])
model.configuration['dict_hdecode'] = os.getcwd() + '/dictionary/dict.hdecode'
model.configuration['dict_hvite'] = os.getcwd() + '/dictionary/dict'
for exp in experiments.keys():
if exp in dirs_done:
experiments[exp].done = True
experiments[exp].launch_options = vars(options)
run_experiments(experiments)
else:
experiments,_ = parse_config([options.recognition_dir + '/config'])
if len(experiments) != 1:
sys.exit("Expected exactly one experiment! (found %s)" % ', '.join(experiments.keys()))
htk.num_tasks = options.num_tasks
htk.default_HERest_pruning = ['300.0', '500.0', '2000.0']
job_runner.default_options["verbosity"] = 1
if current_step >= options.step:
logger.info("Start step: %d (%s)" % (current_step, 'Making reference trn'))
data_manipulation.mlf_to_trn(words_mlf, 'reference.trn', speaker_name_width, ref_del_char)
dicts = []
cur_dict_id = 1
while config.has_option("dict"+str(cur_dict_id), "location"):
dicts.append([config.get("dict"+str(cur_dict_id), "location"),
config.get("dict"+str(cur_dict_id), "prefix"),
config.get("dict"+str(cur_dict_id), "word_suffix")])
cur_dict_id += 1
if len(dicts) > 0:
data_manipulation.import_dictionaries(dicts)
dict = 'dictionary/dict'
dict_hdecode = 'dictionary/dict.hdecode'
baseline_dir = 'baseline'
baseline_lat_dir = baseline_dir + '/lattices.htk'
baseline_lat_dir_rescored = baseline_dir + '/lattices.rescored'
baseline_pass1_mlf = baseline_dir + '/pass1.mlf'
baseline_pass2_mlf = baseline_dir + '/pass2.mlf'
if 'baseline' in experiments:
pass1_trn = baseline_dir + '/pass1.trn'
pass2_trn = baseline_dir + '/pass2.trn'
dest="queue",
help="Fix the queue where things are running in",
default="")
options, configs = parser.parse_args()
if len(options.recognition_dir) == 0:
dirs_done = []
if len(options.dirs_done) > 0:
dirs_done = options.dirs_done.split(',')
experiments, model = parse_config(configs)
model.make_reference()
if model.configuration['dict'] is not None:
data_manipulation.import_dictionaries([
(model.configuration['dict'], '', '')
])
model.configuration['dict_hdecode'] = os.getcwd(
) + '/dictionary/dict.hdecode'
model.configuration['dict_hvite'] = os.getcwd(
) + '/dictionary/dict'
for exp in experiments.keys():
if exp in dirs_done:
experiments[exp].done = True
experiments[exp].launch_options = vars(options)
run_experiments(experiments)
else:
experiments, _ = parse_config([options.recognition_dir + '/config'])
if len(experiments) != 1:
sys.exit("Expected exactly one experiment! (found %s)" %
', '.join(experiments.keys()))