# 静态图使用
import numpy as np
from paddle import fluid

x = fluid.data(name="x", shape=(-1, 3), dtype="float32")
y = fluid.layers.swish(x, beta=2.0)

place = fluid.CPUPlace()
exe = fluid.Executor(place)
start = fluid.default_startup_program()
main = fluid.default_main_program()

data = np.random.randn(2, 3).astype("float32")
exe.run(start)
y_np, = exe.run(main, feed={"x": data}, fetch_list=[y])

data
# array([[-1.1239197 ,  1.3391294 ,  0.03921051],
#        [ 1.1970421 ,  0.02440812,  1.2055548 ]], dtype=float32)
y_np
# array([[-0.2756806 ,  1.0610548 ,  0.01998957],
#        [ 0.9193261 ,  0.01235299,  0.9276883 ]], dtype=float32)

# 动态图使用
import numpy as np
from paddle import fluid
import paddle.fluid.dygraph as dg

data = np.random.randn(2, 3).astype("float32")
place = fluid.CPUPlace()
with dg.guard(place) as g:
    x = dg.to_variable(data)
    y = fluid.layers.swish(x)
    y_np = y.numpy()
data
# array([[-0.0816701 ,  1.1603649 , -0.88325626],
#        [ 0.7522361 ,  1.0978601 ,  0.12987892]], dtype=float32)
y_np
# array([[-0.03916847,  0.8835007 , -0.25835553],
#        [ 0.51126915,  0.82324016,  0.06915068]], dtype=float32)