def open_stream(which_sets=('train', ), port=5557, num_examples=None):
dataset = Blizzard(which_sets=which_sets)
if num_examples == None:
num_examples = dataset.num_examples
data_stream = DataStream.default_stream(dataset,
iteration_scheme=SequentialScheme(
num_examples, batch_size))
data_stream = ScaleAndShift(data_stream,
scale=1 / data_std,
shift=-data_mean / data_std)
data_stream = Mapping(data_stream,
_downsample_and_upsample,
add_sources=('upsampled', ))
data_stream = Mapping(data_stream, _equalize_size)
data_stream = Mapping(data_stream,
_get_residual,
add_sources=('residual', ))
data_stream = FilterSources(data_stream,
sources=(
'upsampled',
'residual',
))
data_stream = Mapping(data_stream, _segment_axis)
data_stream = Mapping(data_stream, _transpose)
data_stream = ForceFloatX(data_stream)
start_server(data_stream, port=port)
def define_stream(which_sets=('train',),
initial_scale=1,
scale=0.5,
batch_size=64,
seq_length=64,
frame_size=128,
tbptt_flag = True,
num_examples=None):
def _segment_axis(data):
# Defined inside so that frame_size is available
x = tuple([numpy.array([segment_axis(x, frame_size, 0) for x in var])
for var in data])
return x
scale = float(scale)
dataset = Blizzard(which_sets=which_sets)
if num_examples is None:
num_examples = batch_size*(dataset.num_examples/batch_size)
data_stream = DataStream.default_stream(
dataset,
iteration_scheme=SequentialScheme(num_examples, batch_size))
data_stream = ScaleAndShift(data_stream,
scale=1/data_std,
shift=-data_mean/float(data_std))
# Original sampling rate
data_stream = Resample(data_stream, scale=initial_scale)
data_stream = Mapping(data_stream, _copy, add_sources=('upsampled',))
data_stream = Resample(data_stream, scale=scale, which_sources=('upsampled',))
data_stream = Resample(data_stream, scale=1/scale, which_sources=('upsampled',))
# data_stream = Mapping(data_stream, _downsample_and_upsample,
# add_sources=('upsampled',))
data_stream = Mapping(data_stream, _equalize_size)
data_stream = Mapping(data_stream, _get_residual,
add_sources=('residual',))
data_stream = FilterSources(data_stream,
sources=('upsampled', 'residual',))
data_stream = Mapping(data_stream, _segment_axis)
data_stream = Mapping(data_stream, _transpose)
return data_stream
# Prepare dataset
#################
def _transpose(data):
return tuple(array.swapaxes(0, 1) for array in data)
data_dir = os.environ['FUEL_DATA_PATH']
data_dir = os.path.join(data_dir, 'blizzard/', 'sp_standardize.npz')
data_stats = numpy.load(data_dir)
sp_mean = data_stats['sp_mean']
sp_std = data_stats['sp_std']
dataset = Blizzard(which_sets=('train', ), filename="sp_blizzard.hdf5")
data_stream = DataStream.default_stream(
dataset,
iteration_scheme=SequentialScheme(
batch_size * (dataset.num_examples / batch_size), batch_size))
data_stream = FilterSources(data_stream, ('sp', ))
data_stream = ScaleAndShift(data_stream,
scale=1 / sp_std,
shift=-sp_mean / sp_std,
which_sources=('sp', ))
data_stream = Mapping(data_stream, _transpose)
data_stream = SegmentSequence(data_stream, seq_size, add_flag=True)
data_stream = ForceFloatX(data_stream)
train_stream = data_stream
num_valid_examples = 4 * 64
from play.datasets.blizzard import Blizzard
import matplotlib
matplotlib.use('Agg')
from matplotlib import pyplot
from fuel.schemes import SequentialScheme
from fuel.streams import DataStream
dataset = Blizzard(which_sets=('test', ), filename="mgc_blizzard.hdf5")
batch_size = 1
data_stream = DataStream.default_stream(dataset,
iteration_scheme=SequentialScheme(
dataset.num_examples, batch_size))
f0, mgc = next(data_stream.get_epoch_iterator())
pyplot.plot(f0[0][:500])
pyplot.savefig('plot_f0.png')
pyplot.close()
pyplot.plot(mgc[0, :500, 3])
pyplot.savefig('plot_mgc.png')
pyplot.close()
pyplot.hist(f0[0, f0[0] > 0])
pyplot.savefig('plot_hist_cond.png')
pyplot.close()
pyplot.hist(mgc.reshape(-1))
def _get_residual(data):
# The order is correct?
ds = numpy.array([x[0]-x[1] for x in zip(*data)])
return (ds,)
data_dir = os.environ['FUEL_DATA_PATH']
data_dir = os.path.join(data_dir, 'blizzard/', 'blizzard_standardize.npz')
data_stats = numpy.load(data_dir)
data_mean = data_stats['data_mean']
data_std = data_stats['data_std']
which_sets= ('test',)
dataset = Blizzard(which_sets = which_sets)
data_stream = DataStream.default_stream(
dataset, iteration_scheme=SequentialScheme(
dataset.num_examples, batch_size))
epoch_iterator = data_stream.get_epoch_iterator()
raw_audio = next(epoch_iterator)[0]
for i in xrange(n_iter-1):
x_tr = next(epoch_iterator)[0]
raw_audio = numpy.hstack([raw_audio, x_tr])
job_id = sys.argv[1]
save_dir = os.environ['RESULTS_DIR']
exp_path = os.path.join(save_dir,'blizzard/', job_id + "/")
return tuple(array.swapaxes(0, 1) for array in data)
def _segment_axis(data):
x = numpy.array([segment_axis(x, frame_size, 0) for x in data[0]])
return (x, )
data_dir = os.environ['FUEL_DATA_PATH']
data_dir = os.path.join(data_dir, 'blizzard/', 'blizzard_standardize.npz')
data_stats = numpy.load(data_dir)
data_mean = data_stats['data_mean']
data_std = data_stats['data_std']
dataset = Blizzard(which_sets=('train', ))
data_stream = DataStream.default_stream(dataset,
iteration_scheme=SequentialScheme(
dataset.num_examples, batch_size))
data_stream = ScaleAndShift(data_stream,
scale=1 / data_std,
shift=-data_mean / data_std)
data_stream = Mapping(data_stream, _segment_axis)
data_stream = Mapping(data_stream, _transpose)
data_stream = ForceFloatX(data_stream)
train_stream = data_stream
num_valid_examples = 4 * 64 * 5
dataset = Blizzard(which_sets=('valid', ))
data_stream = DataStream.default_stream(dataset,
iteration_scheme=SequentialScheme(
def _zero_for_unvoiced(data):
#Multiply by 0 the unvoiced components. Hardcoded.
return tuple([data[0] * data[3], data[1], data[2], data[3]])
data_dir = os.environ['FUEL_DATA_PATH']
data_dir = os.path.join(data_dir, 'blizzard/', 'sp_standardize.npz')
data_stats = numpy.load(data_dir)
sp_mean = data_stats['sp_mean']
sp_std = data_stats['sp_std']
f0_mean = data_stats['f0_mean']
f0_std = data_stats['f0_std']
dataset = Blizzard(which_sets=('test', ), filename="sp_blizzard_80h_phon.hdf5")
data_stream = DataStream.default_stream(
dataset,
iteration_scheme=SequentialScheme(
batch_size * (dataset.num_examples / batch_size), batch_size))
epoch_iterator = data_stream.get_epoch_iterator()
# data_stream = Mapping(data_stream, _is_nonzero, add_sources = ('voiced',))
# data_stream = ScaleAndShift(data_stream,
# scale = 1/sp_std,
# shift = -sp_mean/sp_std,
# which_sources = ('sp',))
# data_stream = ScaleAndShift(data_stream,
# scale = 1/f0_std,
mgc = np.apply_along_axis(SPTK.mgcep, 1, frames, order, alpha, gamma)
mgc_sp = np.apply_along_axis(SPTK.mgc2sp, 1, mgc, alpha, gamma,
frame_length).real
mgc_sp_test = np.hstack([mgc_sp, mgc_sp[:, ::-1][:, 1:-1]])
mgc_sp_test = mgc_sp_test.copy(order='C')
# Check in original data, that the processing was good
import h5py
import fuel
from play.datasets.blizzard import Blizzard
from fuel.schemes import SequentialScheme
from fuel.streams import DataStream
dataset_mgc = Blizzard(which_sets=('train', 'valid', 'test'),
filename="mgc_blizzard.hdf5")
dataset = Blizzard(which_sets=('train', 'valid', 'test'))
batch_size = 2
data_stream = DataStream.default_stream(dataset,
iteration_scheme=SequentialScheme(
dataset.num_examples, batch_size))
data_stream_mgc = DataStream.default_stream(dataset_mgc,
iteration_scheme=SequentialScheme(
dataset_mgc.num_examples,
batch_size))
raw = next(data_stream.get_epoch_iterator())[0]
f0, mgc = next(data_stream_mgc.get_epoch_iterator())