lr_decay = (final_lr / lr)**(1. / n_epoch)
# layer's parameters
n_inputs = 27
n_units = 2000
n_classes = 27
n_hidden_layer = 1
# binary/ternary weights for LSTM
binary_training = True
ternary_training = False
stochastic_training = True
print('Preparing the dataset')
train_set = get_stream(which_set="train",
batch_size=batch_size,
length=length,
augment=True)
valid_set = get_stream(which_set="valid",
batch_size=batch_size,
length=length)
test_set = get_stream(which_set="test",
batch_size=batch_size,
length=length)
print('Creating the model')
class create_model(Network):
def __init__(self, rng):
Network.__init__(self,
n_hidden_layer=n_hidden_layer,
final_lr = 0.001
lr_decay = (final_lr / lr)**(1. / n_epoch)
# layer's parameters
n_inputs = 1
n_units = 100
n_classes = 10
n_hidden_layer = 1
# binary/ternary weights for LSTM
binary_training = False
ternary_training = False
stochastic_training = True
print('Preparing the dataset')
train_set = get_stream(which_set="train", batch_size=batch_size)
valid_set = get_stream(which_set="valid", batch_size=batch_size)
test_set = get_stream(which_set="test", batch_size=batch_size)
print('Creating the model')
class create_model(Network):
def __init__(self, rng):
Network.__init__(self,
n_hidden_layer=n_hidden_layer,
BN_LSTM=BN_LSTM,
length=length)
print("LSTM layer:")
self.layer.append(