if b_init_args is None:
b_init_args = {}
Layer.__init__(self, prev_layer=prev_layer, name=name)self.inputs = prev_layer.outputs
if self.inputs.get_shape().ndims != 2:
raise Exception("The input dimension must be rank 2, please reshape or flatten it")
if use_gemm:
raise Exception("TODO. The current version use tf.matmul for inferencing.")
n_in = int(self.inputs.get_shape()[-1])
self.n_units = n_units
logging.info("BinaryDenseLayer %s: %d %s" % (self.name, self.n_units, act.__name__))
with tf.variable_scope(name):
W = tf.get_variable(name="W", shape=(n_in, n_units), initializer=W_init, dtype=LayersConfig.tf_dtype, **W_init_args)
// W = tl.act.sign(W) // dont update ...
W = quantize(W)
// W = tf.Variable(W)
// print(W)
if b_init is not None:
try:
b = tf.get_variable(name="b", shape=(n_units), initializer=b_init, dtype=LayersConfig.tf_dtype, **b_init_args)
except Exception: // If initializer is a constant, do not specify shape.
b = tf.get_variable(name="b", initializer=b_init, dtype=LayersConfig.tf_dtype, **b_init_args)
self.outputs = act(tf.matmul(self.inputs, W) + b)
// self.outputs = act(xnor_gemm(self.inputs, W) + b) // TODO
else:
self.outputs = act(tf.matmul(self.inputs, W))
// self.outputs = act(xnor_gemm(self.inputs, W)) // TODO
self.all_layers.append(self.outputs)
After Change
b_init_args=None,
name="binary_dense",
):
super(BinaryDenseLayer, self).__init__(prev_layer=prev_layer, name=name)
logging.info("BinaryDenseLayer %s: %d %s" % (name, n_units, act.__name__))
self.inputs = prev_layer.outputs
if W_init_args is None:
W_init_args = {}
if b_init_args is None:
b_init_args = {}
if self.inputs.get_shape().ndims != 2:
raise Exception("The input dimension must be rank 2, please reshape or flatten it")
if use_gemm:
raise Exception("TODO. The current version use tf.matmul for inferencing.")
n_in = int(self.inputs.get_shape()[-1])
self.n_units = n_units
with tf.variable_scope(name):
W = tf.get_variable(name="W", shape=(n_in, n_units), initializer=W_init, dtype=LayersConfig.tf_dtype, **W_init_args)
// W = tl.act.sign(W) // dont update ...
W = quantize(W)
// W = tf.Variable(W)
// print(W)
if b_init is not None:
try:
b = tf.get_variable(name="b", shape=(n_units), initializer=b_init, dtype=LayersConfig.tf_dtype, **b_init_args)
except Exception: // If initializer is a constant, do not specify shape.
b = tf.get_variable(name="b", initializer=b_init, dtype=LayersConfig.tf_dtype, **b_init_args)
self.outputs = act(tf.matmul(self.inputs, W) + b)
// self.outputs = act(xnor_gemm(self.inputs, W) + b) // TODO
else:
self.outputs = act(tf.matmul(self.inputs, W))
// self.outputs = act(xnor_gemm(self.inputs, W)) // TODO
self.all_layers.append(self.outputs)
