|
|
|
@ -1909,39 +1909,34 @@ class TestBook(LayerTest):
|
|
|
|
|
return (out)
|
|
|
|
|
|
|
|
|
|
def test_kldiv_loss(self):
|
|
|
|
|
program = Program()
|
|
|
|
|
with program_guard(program):
|
|
|
|
|
with program_guard(fluid.default_main_program(),
|
|
|
|
|
fluid.default_startup_program()):
|
|
|
|
|
x = layers.data(name='x', shape=[32, 128, 128], dtype="float32")
|
|
|
|
|
target = layers.data(
|
|
|
|
|
name='target', shape=[32, 128, 128], dtype="float32")
|
|
|
|
|
loss = layers.kldiv_loss(x=x, target=target, reduction='batchmean')
|
|
|
|
|
self.assertIsNotNone(loss)
|
|
|
|
|
|
|
|
|
|
print(str(program))
|
|
|
|
|
return (loss)
|
|
|
|
|
|
|
|
|
|
def test_temporal_shift(self):
|
|
|
|
|
program = Program()
|
|
|
|
|
with program_guard(program):
|
|
|
|
|
with program_guard(fluid.default_main_program(),
|
|
|
|
|
fluid.default_startup_program()):
|
|
|
|
|
x = layers.data(name="X", shape=[16, 4, 4], dtype="float32")
|
|
|
|
|
out = layers.temporal_shift(x, seg_num=4, shift_ratio=0.2)
|
|
|
|
|
self.assertIsNotNone(out)
|
|
|
|
|
print(str(program))
|
|
|
|
|
return (out)
|
|
|
|
|
|
|
|
|
|
def test_shuffle_channel(self):
|
|
|
|
|
program = Program()
|
|
|
|
|
with program_guard(program):
|
|
|
|
|
with program_guard(fluid.default_main_program(),
|
|
|
|
|
fluid.default_startup_program()):
|
|
|
|
|
x = layers.data(name="X", shape=[16, 4, 4], dtype="float32")
|
|
|
|
|
out = layers.shuffle_channel(x, group=4)
|
|
|
|
|
self.assertIsNotNone(out)
|
|
|
|
|
print(str(program))
|
|
|
|
|
return (out)
|
|
|
|
|
|
|
|
|
|
def test_pixel_shuffle(self):
|
|
|
|
|
program = Program()
|
|
|
|
|
with program_guard(program):
|
|
|
|
|
with program_guard(fluid.default_main_program(),
|
|
|
|
|
fluid.default_startup_program()):
|
|
|
|
|
x = layers.data(name="X", shape=[9, 4, 4], dtype="float32")
|
|
|
|
|
out = layers.pixel_shuffle(x, upscale_factor=3)
|
|
|
|
|
self.assertIsNotNone(out)
|
|
|
|
|
print(str(program))
|
|
|
|
|
return (out)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
|