def test_elementwise_binary(self):
desc_a = cutensor.create_tensor_descriptor(self.a)
desc_c = cutensor.create_tensor_descriptor(self.c)
d = cutensor.elementwise_binary(self.alpha, self.a, desc_a,
self.mode_a, self.gamma, self.c,
desc_c, self.mode_c)
assert d.dtype == self.dtype
testing.assert_allclose(self.alpha * self.a_transposed +
self.gamma * self.c_transposed,
d,
rtol=self.tol,
atol=self.tol)
def test_elementwise_binary(self):
desc_a = cutensor.create_tensor_descriptor(self.a, ct.OP_SIGMOID)
desc_c = cutensor.create_tensor_descriptor(self.c, ct.OP_ABS)
d = cutensor.elementwise_binary(
self.alpha, self.a, desc_a, self.mode_a,
self.gamma, self.c, desc_c, self.mode_c,
op_AC=ct.OP_MUL
)
testing.assert_allclose(
self.alpha * (1 / (1 + cupy.exp(-self.a_transposed))) *
self.gamma * cupy.abs(self.c),
d,
rtol=1e-6, atol=1e-6
)
def test_elementwise_binary(self):
desc_a = cutensor.create_tensor_descriptor(self.a)
desc_c = cutensor.create_tensor_descriptor(self.c)
d = cutensor.elementwise_binary(
self.alpha, self.a, desc_a, self.mode_a,
self.gamma, self.c, desc_c, self.mode_c
)
assert d.dtype == numpy.float32
testing.assert_allclose(
self.alpha.item() * self.a_transposed +
self.gamma.item() * self.c_transposed,
d,
rtol=1e-6, atol=1e-6
)
def test_elementwise_binary_out(self):
out = testing.shaped_random(
(30, 20, 40), cupy, self.dtype, seed=3)
desc_a = cutensor.create_tensor_descriptor(self.a)
desc_c = cutensor.create_tensor_descriptor(self.c)
d = cutensor.elementwise_binary(
self.alpha, self.a, desc_a, self.mode_a,
self.gamma, self.c, desc_c, self.mode_c, out=out
)
assert d is out
testing.assert_allclose(
self.alpha * self.a_transposed +
self.gamma * self.c_transposed,
d,
rtol=self.tol, atol=self.tol
)
extent = {'x': 400, 'y': 200, 'z': 300}
a = cupy.random.random([extent[i] for i in mode_a])
c = cupy.random.random([extent[i] for i in mode_c])
a = a.astype(dtype)
c = c.astype(dtype)
desc_a = cutensor.create_tensor_descriptor(a)
desc_c = cutensor.create_tensor_descriptor(c)
alpha = 1.1
gamma = 1.3
# rehearsal
d = cutensor.elementwise_binary(alpha, a, desc_a, mode_a, gamma, c, desc_c,
mode_c)
ev_start = stream.Event()
ev_end = stream.Event()
st = stream.Stream()
with st:
# measurement
ev_start.record()
d = cutensor.elementwise_binary(alpha, a, desc_a, mode_a, gamma, c, desc_c,
mode_c)
ev_end.record()
st.synchronize()
elapsed_ms = stream.get_elapsed_time(ev_start, ev_end)
transfer_byte = d.size * d.itemsize
if alpha != 0.0:
