def softmax(logits, mask=None, scope=None):
with tf.name_scope(scope or "Softmax"):
if mask is not None:
logits = exp_mask(logits, mask)
flat_logits = flatten(logits, 1)
flat_out = tf.nn.softmax(flat_logits)
out = reconstruct(flat_out, logits, 1)
return out
def linear(args,
output_size,
bias,
bias_start=0.0,
scope=None,
squeeze=False,
wd=0.0,
input_keep_prob=1.0,
is_train=None): #, name_w='', name_b=''
# if args is None or (nest.is_sequence(args) and not args):
# raise ValueError("`args` must be specified")
# if not nest.is_sequence(args):
# args = [args]
flat_args = [flatten(arg, 1) for arg in args]
# if input_keep_prob < 1.0:
# assert is_train is not None
flat_args = [tf.nn.dropout(arg, input_keep_prob) for arg in flat_args]
total_arg_size = 0
shapes = [a.get_shape() for a in flat_args]
for shape in shapes:
if shape.ndims != 2:
raise ValueError("linear is expecting 2D arguments: %s" % shapes)
if shape[1].value is None:
raise ValueError(
"linear expects shape[1] to be provided for shape %s, "
"but saw %s" % (shape, shape[1]))
else:
total_arg_size += shape[1].value
# print(total_arg_size)
# exit()
dtype = [a.dtype for a in flat_args][0]
# scope = tf.get_variable_scope()
with tf.variable_scope(scope) as outer_scope:
weights = tf.get_variable(_WEIGHTS_VARIABLE_NAME,
[total_arg_size, output_size],
dtype=dtype)
if len(flat_args) == 1:
res = tf.matmul(flat_args[0], weights)
else:
res = tf.matmul(tf.concat(flat_args, 1), weights)
if not bias:
flat_out = res
else:
with tf.variable_scope(outer_scope) as inner_scope:
inner_scope.set_partitioner(None)
biases = tf.get_variable(_BIAS_VARIABLE_NAME, [output_size],
dtype=dtype,
initializer=tf.constant_initializer(
bias_start, dtype=dtype))
flat_out = tf.nn.bias_add(res, biases)
out = reconstruct(flat_out, args[0], 1)
if squeeze:
out = tf.squeeze(out, [len(args[0].get_shape().as_list()) - 1])
if wd:
add_wd(wd)
return out
out = conv1d(in_,
filter_size,
height,
padding,
keep_prob=keep_prob,
scope="conv1d_{}".format(height))
outs.append(out)
concat_out = tf.concat(outs, axis=2)
return concat_out
if __name__ == '__main__':
a = tf.Variable(np.random.random(size=(2, 2, 4)))
b = tf.Variable(np.random.random(size=(2, 3, 4)))
c = tf.tile(tf.expand_dims(a, 2), [1, 1, 3, 1])
test = flatten(c, 1)
out = reconstruct(test, c, 1)
d = tf.tile(tf.expand_dims(b, 1), [1, 2, 1, 1])
e = linear(
[c, d, c * d],
1,
bias=False,
scope="test",
)
# f = softsel(d, e)
with tf.Session() as sess:
tf.global_variables_initializer().run()
print(sess.run(test))
print(sess.run(tf.shape(out)))
exit()
print(sess.run(tf.shape(a)))