def square_dist(self, X, Xs):
X = aet.mul(X, 1.0 / self.ls)
X2 = aet.sum(aet.square(X), 1)
if Xs is None:
sqd = -2.0 * aet.dot(X, aet.transpose(X)) + (
aet.reshape(X2, (-1, 1)) + aet.reshape(X2, (1, -1)))
else:
Xs = aet.mul(Xs, 1.0 / self.ls)
Xs2 = aet.sum(aet.square(Xs), 1)
sqd = -2.0 * aet.dot(X, aet.transpose(Xs)) + (
aet.reshape(X2, (-1, 1)) + aet.reshape(Xs2, (1, -1)))
return aet.clip(sqd, 0.0, np.inf)
def __call__(self, X):
XY = X.dot(X.T)
x2 = at.sum(X**2, axis=1).dimshuffle(0, "x")
X2e = at.repeat(x2, X.shape[0], axis=1)
H = X2e + X2e.T - 2.0 * XY
V = at.sort(H.flatten())
length = V.shape[0]
# median distance
m = at.switch(
at.eq((length % 2), 0),
# if even vector
at.mean(V[((length // 2) - 1):((length // 2) + 1)]),
# if odd vector
V[length // 2],
)
h = 0.5 * m / at.log(floatX(H.shape[0]) + floatX(1))
# RBF
Kxy = at.exp(-H / h / 2.0)
# Derivative
dxkxy = -at.dot(Kxy, X)
sumkxy = at.sum(Kxy, axis=-1, keepdims=True)
dxkxy = at.add(dxkxy, at.mul(X, sumkxy)) / h
return Kxy, dxkxy
def test_composite_elemwise_float16(self):
w = aesara.tensor.bvector()
x = aesara.tensor.vector(dtype="float16")
y = aesara.tensor.fvector()
cz = tensor.tanh(x + tensor.cast(y, "float16"))
o = (cz - cz**2 + tensor.cast(x, "int16") + tensor.cast(x, "float32") +
tensor.cast(w, "float16") - tensor.constant(np.float16(1.0)))
aesara.function([w, x, y], o, mode=mode_with_gpu)
v = aesara.tensor.vector(dtype="uint8")
w = aesara.tensor.vector(dtype="float16")
x = aesara.tensor.vector(dtype="float16")
y = aesara.tensor.vector(dtype="float16")
z = aesara.tensor.vector(dtype="float16")
o = tensor.switch(v, tensor.mul(w, x, y), z)
aesara.function([v, w, x, y, z], o, mode=mode_with_gpu)
def __call__(self, X):
return at.mul(self.m1(X), self.m2(X))