mirrors/DeepSpeech
2
0
Fork 0
mirror of https://github.com/mozilla/DeepSpeech.git synced 2025-10-26 11:19:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #102 from mozilla/issue101

Browse Source 
Fixes #101
This commit is contained in:
Kelly Davis 2016-10-27 17:02:45 +02:00 committed by GitHub
commit 47bb4babde
2 changed files with 48 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
util/importers/librivox.py
View File

@ -73,25 +73,30 @@ class DataSet(object):
def _populate_batch_queue(self):
with self._graph.as_default():
while True:
n_steps = 0
sources = []
targets = []
for index, (txt_file, wav_file) in enumerate(self._files_circular_list):
if index >= self._batch_size:
break
next_source = audiofile_to_input_vector(wav_file, self._numcep, self._numcontext)
if n_steps < next_source.shape[0]:
n_steps = next_source.shape[0]
sources.append(next_source)
with open(txt_file) as open_txt_file:
targets.append(open_txt_file.read())
target = texts_to_sparse_tensor(targets)
for index, next_source in enumerate(sources):
npad = ((0,(n_steps - next_source.shape[0])), (0,0))
sources[index] = np.pad(next_source, pad_width=npad, mode='constant')
source = np.array(sources)
self._batch_queue.put((source, target))
n_steps = 0
sources = []
targets = []
batch_index = 0
for txt_file, wav_file in self._files_circular_list:
if batch_index == self._batch_size:
# Put batch on queue
target = texts_to_sparse_tensor(targets)
for index, next_source in enumerate(sources):
npad = ((0,(n_steps - next_source.shape[0])), (0,0))
sources[index] = np.pad(next_source, pad_width=npad, mode='constant')
source = np.array(sources)
self._batch_queue.put((source, target))
n_steps = 0
sources = []
targets = []
batch_index = 0
next_source = audiofile_to_input_vector(wav_file, self._numcep, self._numcontext)
if n_steps < next_source.shape[0]:
n_steps = next_source.shape[0]
sources.append(next_source)
with open(txt_file) as open_txt_file:
targets.append(open_txt_file.read())
batch_index = batch_index + 1
def next_batch(self):
source, target = self._batch_queue.get()

43
util/importers/ted_lium.py
View File

@ -79,25 +79,30 @@ class DataSet(object):
def _populate_batch_queue(self):
with self._graph.as_default():
while True:
n_steps = 0
sources = []
targets = []
for index, (txt_file, wav_file) in enumerate(self._files_circular_list):
if index >= self._batch_size:
break
next_source = audiofile_to_input_vector(wav_file, self._numcep, self._numcontext)
if n_steps < next_source.shape[0]:
n_steps = next_source.shape[0]
sources.append(next_source)
with open(txt_file) as open_txt_file:
targets.append(open_txt_file.read())
target = texts_to_sparse_tensor(targets)
for index, next_source in enumerate(sources):
npad = ((0,(n_steps - next_source.shape[0])), (0,0))
sources[index] = np.pad(next_source, pad_width=npad, mode='constant')
source = np.array(sources)
self._batch_queue.put((source, target))
n_steps = 0
sources = []
targets = []
batch_index = 0
for txt_file, wav_file in self._files_circular_list:
if batch_index == self._batch_size:
# Put batch on queue
target = texts_to_sparse_tensor(targets)
for index, next_source in enumerate(sources):
npad = ((0,(n_steps - next_source.shape[0])), (0,0))
sources[index] = np.pad(next_source, pad_width=npad, mode='constant')
source = np.array(sources)
self._batch_queue.put((source, target))
n_steps = 0
sources = []
targets = []
batch_index = 0
next_source = audiofile_to_input_vector(wav_file, self._numcep, self._numcontext)
if n_steps < next_source.shape[0]:
n_steps = next_source.shape[0]
sources.append(next_source)
with open(txt_file) as open_txt_file:
targets.append(open_txt_file.read())
batch_index = batch_index + 1
def next_batch(self):
source, target = self._batch_queue.get()