def model(th):
assert isinstance(th, m.Config)
model = m.get_container(th, flatten=False)
model.add(m.Conv2D(64, kernel_size=3, strides=1))
model.add(m.Activation('relu'))
model.add(m.MaxPool2D(2, strides=2))
model.add(m.Conv2D(192, kernel_size=3, strides=1))
model.add(m.Activation('relu'))
model.add(m.MaxPool2D(2, strides=2))
for filters in (384, 256, 256):
model.add(m.Conv2D(filters, kernel_size=3))
model.add(m.Activation('relu'))
model.add(m.MaxPool2D(3, strides=2))
model.add(m.Flatten())
for dim in (2048, 2048):
if th.dropout > 0: model.add(m.Dropout(1. - th.dropout))
model.add(m.Linear(dim))
model.add(m.Activation('relu'))
return m.finalize(th, model)
def model(th):
assert isinstance(th, m.Config)
model = m.get_container(th, flatten=True)
model.add(Dense(th.layer_width, th.spatial_activation))
model.add(Highway(
th.layer_width, th.num_layers, th.spatial_activation,
t_bias_initializer=th.bias_initializer))
# model.register_extractor(m.LinearHighway.extractor)
return m.finalize(th, model)
def model(th):
assert isinstance(th, m.Config)
model = m.get_container(th, flatten=True)
for dim in th.fc_dims:
model.add(m.Linear(dim))
if th.use_batchnorm: model.add(m.BatchNormalization())
model.add(m.Activation(th.spatial_activation))
if th.dropout > 0:
model.add(m.Dropout(train_keep_prob=1. - th.dropout))
return m.finalize(th, model)
def model(th):
assert isinstance(th, m.Config)
model = m.get_container(th, flatten=True)
model.add(Dense(th.layer_width, th.spatial_activation))
model.add(
SLHighway(
config_string=th.group_string,
num_layers=th.num_layers,
head_size=th.head_size,
activation=th.spatial_activation,
gutter=th.gutter,
gutter_bias=th.gutter_bias,
))
# model.register_extractor(m.LinearHighway.extractor)
return m.finalize(th, model)