def test_linalg_norm():
"""All that I'm asking of this test is that it doesn't error out."""
x = cp.array([1, 2, 3])
l2 = cp.linalg.norm(x)
def loss(w, x, y):
preds = model(w, x)
loss_score = cp.mean(cp.power(preds.ravel() - y.ravel(), cp.array(2)))
return loss_score
def fun(x):
return cp.array([[2 * x, x + 1], [x, x]])
def test_sparse_coo_matrix(eye):
"""This just has to not error out."""
data = cp.array([1, 2, 3]).astype("float32")
rows = cp.array([1, 2, 3]).astype("float32")
cols = cp.array([1, 3, 4]).astype("float32")
print(eye.shape)
def fun(x):
return cp.array(x)
def fun(x):
return cp.array([x, x])
def fun(x):
A = cp.array([x, x * 1.0, x + 2.5])
return A + A
def fun(x):
A = cp.array([[x, x * 1.0, x + 2.5], [x**2, x, x / 2.0]])
return A + A
def fun(x):
# return cp.array([x, x * 1.0, x + 2.5])
return cp.array([x, x, x])
def fun(x):
return cp.array([[x, x * 1.0, x + 2.5], [x**2, x, x / 2.0]])
def test_simple_append_arr():
A = cp.array([1.0, 2.0, 3.0])
b = 4.0
check_grads(cp.append, argnum=(0, 1))(A, b)
def fun(x):
return cp.maximum(cp.array([x, x, 0.5]),
cp.array([[x, 0.5, x], [x, x, 0.5]]))
def fun(x):
return cp.min(
cp.array([[x, x, x], [x, 0.5, 0.5], [0.5, 0.5, 0.5], [x, x, 0.5]]),
axis=0,
)
def fun(x):
return cp.max(cp.array([[x, x], [x, 0.5]]), axis=1)
def test_nan_to_num():
y = cp.array([0.0, cp.nan, cp.inf, -cp.inf])
fun = lambda x: cp.sum(cp.sin(cp.nan_to_num(x + y)))
x = cp.random.randn(4)
check_grads(fun)(x)