def __getitem__(self, index):
x_batch = self.X[index]
y_batch = self.Y[index]
if self.run_mockingjay:
x_batch = process_test_MAM_data(spec=(x_batch, ),
config=self.mock_config)
return x_batch, y_batch
def __getitem__(self, index):
# Load acoustic feature and pad
x_batch = [
torch.FloatTensor(np.load(os.path.join(self.root, x_file)))
for x_file in self.X[index]
]
x_pad_batch = pad_sequence(x_batch, batch_first=True)
# Return (x_spec, speaker_label)
s_batch = torch.LongTensor([
self.speaker2idx[self.get_speaker_from_path(x_file)]
for x_file in self.X[index]
])
if self.run_mockingjay:
x_pad_batch = process_test_MAM_data(spec=(x_pad_batch, ),
config=self.mock_config)
return x_pad_batch, s_batch
def __getitem__(self, index):
# Load acoustic feature and pad
x_batch = [
torch.FloatTensor(np.load(os.path.join(self.root, x_file)))
for x_file in self.X[index]
]
x_pad_batch = pad_sequence(x_batch, batch_first=True)
p_batch = [torch.LongTensor(pickle.load(open(os.path.join(self.phone_path, \
x_file.replace('npy', 'pkl')), "rb"))) for x_file in self.X[index]]
p_pad_batch = pad_sequence(p_batch, batch_first=True)
x_match_batch, p_match_batch = self.match_sequence(
x_pad_batch, p_pad_batch)
# Return (x_spec, phone_label)
if self.run_mockingjay:
x_match_batch = process_test_MAM_data(spec=(x_match_batch, ),
config=self.mock_config)
return x_match_batch, p_match_batch
def __getitem__(self, index):
# Load acoustic feature and pad
x_batch = [
torch.FloatTensor(np.load(os.path.join(self.root, 'npy', x_file)))
for x_file in self.X[index]
] # [(seq, feature), ...]
x_pad_batch = pad_sequence(
x_batch, batch_first=True
) # (batch, seq, feature) with all seq padded with zeros to align the longest seq in this batch
seq_len = x_pad_batch.size(1)
x_pad_batch = x_pad_batch[:,
torch.arange(0, seq_len, self.
config['sample_rate']), :]
# Load label
y_batch = torch.LongTensor(self.Y[index]) # (batch, )
# y_broadcast_int_batch = y_batch.repeat(x_pad_batch.size(1), 1).T # (batch, seq)
if self.run_mockingjay:
x_pad_batch = process_test_MAM_data(spec=(x_pad_batch, ),
config=self.mock_config)
return x_pad_batch, y_batch
def __getitem__(self, index):
# Load acoustic feature and pad
x_batch = [
torch.FloatTensor(np.load(os.path.join(self.npy_dir, x_file)))
for x_file in self.X[index]
] # [(seq, feature), ...]
x_pad_batch = pad_sequence(
x_batch, batch_first=True
) # (batch, seq, feature) with all seq padded with zeros to align the longest seq in this batch
truncate_length = self.config['truncate_length']
if x_pad_batch.size(1) > self.config['truncate_length']:
x_pad_batch = x_pad_batch[:, :truncate_length, :]
# Load label
if self.config['label_mode'] == 'regression':
y_batch = torch.FloatTensor(self.Y[index]) # (batch, )
else:
y_batch = torch.LongTensor(self.Y[index]) # (batch, )
# y_broadcast_int_batch = y_batch.repeat(x_pad_batch.size(1), 1).T # (batch, seq)
if self.run_mockingjay:
x_pad_batch = process_test_MAM_data(spec=(x_pad_batch, ),
config=self.mock_config)
return x_pad_batch, y_batch
