def train(opts, dirs):
## Handle corpus:
print(" -- Gather corpus")
## Get names of directories containing corpus data (all txt and wav):
corpora = []
if opts.command_line_corpus:
for location in opts.command_line_corpus:
assert os.path.isdir(location)
corpora.append(location)
else:
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.SPEAKERS,
opts.speaker, "txt"))
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.SPEAKERS,
opts.speaker, "wav"))
# additional large text corpus:
if opts.text_corpus_name:
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.TEXT_CORPORA,
opts.text_corpus_name))
# Set file number
if opts.file_num:
file_num = int(opts.file_num)
else:
file_num = float("inf")
# Get names of individual txt and wav files:
voice_data = []
for c in corpora:
count = 0
file_list = sorted(os.listdir(c))
if opts.shuffle:
random.seed(1)
random.shuffle(file_list)
for f in file_list:
if '._' not in f:
voice_data.append(os.path.join(c, f))
count += 1
# Stop appending voice data at file_num
if count >= file_num:
break
corpus = Corpus.Corpus(voice_data)
print(" -- Train voice")
voice = Voice(opts.speaker, opts.lang, opts.config, opts.stage, \
dirs, clear_old_data=opts.clear, max_cores=opts.max_cores)
## Train the voice (i.e. train processors in pipeline context):
voice.train(corpus)
def train(opts, dirs):
## Handle corpus:
print " -- Gather corpus"
## Get names of directories containing corpus data (all txt and wav):
corpora = []
if opts.command_line_corpus:
for location in opts.command_line_corpus:
assert os.path.isdir(location)
corpora.append(location)
else:
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.SPEAKERS,
opts.speaker, "txt"))
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.SPEAKERS,
opts.speaker, "wav"))
# additional large text corpus:
if opts.text_corpus_name:
corpora.append(
os.path.join(dirs['CORPUS'], opts.lang, fname.TEXT_CORPORA,
opts.text_corpus_name))
## Get names of individual txt and wav files:
voice_data = []
for c in corpora:
for f in os.listdir(c):
voice_data.append(os.path.join(c, f))
corpus = Corpus.Corpus(voice_data)
print " -- Train voice"
voice = Voice(opts.speaker, opts.lang, opts.config, opts.stage, \
dirs, clear_old_data=opts.clear, max_cores=opts.max_cores)
## Train the voice (i.e. train processors in pipeline context):
voice.train(corpus)