#######################
nb_sequences = 32
######################
# Load test database #
######################
print('Loading test datatable...', end='')
(src_test_datatable,
src_test_masks,
trg_test_datatable,
trg_test_masks,
max_test_length,
test_speakers_max,
test_speakers_min
) = s2s.seq2seq2_load_datatable(
'data/seq2seq_test_datatable.h5'
)
print('done')
#############################
# Load model and parameters #
#############################
with h5py.File('training_results/seq2seq_training_params.h5', 'r') as f:
epochs = f.attrs.get('epochs')
learning_rate = f.attrs.get('learning_rate')
optimizer = f.attrs.get('optimizer')
loss = f.attrs.get('loss')
train_speakers_max = f.attrs.get('train_speakers_max')
train_speakers_min = f.attrs.get('train_speakers_min')
print('Re-initializing model')
'data/seq2seq_test_datatable'
)
print('done')
else:
# Retrieve datatables from .h5 files
print('Loading training datatable...', end='')
(src_train_datatable,
src_train_masks,
trg_train_datatable,
trg_train_masks,
max_train_length,
train_speakers_max,
train_speakers_min
) = s2s.seq2seq2_load_datatable(
'data/seq2seq_train_datatable.h5'
)
print('done')
##################
# Normalize data #
##################
# Iterate over sequence 'slices'
assert src_train_datatable.shape[0] == trg_train_datatable.shape[0]
for i in range(src_train_datatable.shape[0]):
(
src_train_datatable[i, :, 0:42],
trg_train_datatable[i, :, 0:42]
) = maxmin_scaling(
src_train_datatable[i, :, :],