def test_cross_entropy_with_softmax():
data1_shape = (1, 2)
label1_shape = (1, )
data2_shape = (1, 3)
label2_shape = (1, )
data1 = np.array([1, 0.5], dtype=np.float32).reshape(data1_shape)
label1 = np.array([1], dtype=np.int32).reshape(label1_shape)
expect1 = F.cross_entropy(F.softmax(tensor(data1)), tensor(label1)).numpy()
data2 = np.array([0.3, 0.4, 0.3], dtype=np.float32).reshape(data2_shape)
label2 = np.array([1], dtype=np.int32).reshape(label2_shape)
expect2 = F.cross_entropy(F.softmax(tensor(data2)), tensor(label2)).numpy()
cases = [
{
"input": [data1, label1],
"output": expect1,
},
{
"input": [data2, label2],
"output": expect2,
},
]
opr_test(cases, F.cross_entropy_with_softmax)
def test_binary_cross_entropy():
data1_shape = (2, 2)
label1_shape = (2, 2)
data2_shape = (2, 3)
label2_shape = (2, 3)
def sigmoid(x):
return 1 / (1 + np.exp(-x))
def compare_fn(x, y):
assertTensorClose(x.numpy(), y, max_err=5e-4)
np.random.seed(123)
data1 = sigmoid(np.random.uniform(size=data1_shape).astype(np.float32))
label1 = np.random.uniform(size=label1_shape).astype(np.float32)
expect1 = np.array([0.6361], dtype=np.float32)
np.random.seed(123)
data2 = sigmoid(np.random.uniform(size=data2_shape).astype(np.float32))
label2 = np.random.uniform(size=label2_shape).astype(np.float32)
expect2 = np.array([0.6750], dtype=np.float32)
cases = [
{
"input": [data1, label1],
"output": expect1,
},
{
"input": [data2, label2],
"output": expect2,
},
]
opr_test(cases, F.binary_cross_entropy, compare_fn=compare_fn)
def test_cross_entropy():
data1_shape = (1, 2)
label1_shape = (1, )
data2_shape = (1, 3)
label2_shape = (1, )
data1 = np.array([0.5, 0.5], dtype=np.float32).reshape(data1_shape)
label1 = np.array([1], dtype=np.int32).reshape(label1_shape)
expect1 = np.array([-np.log(0.5)], dtype=np.float32)
data2 = np.array([0.3, 0.4, 0.3], dtype=np.float32).reshape(data2_shape)
label2 = np.array([1], dtype=np.int32).reshape(label2_shape)
expect2 = np.array([-np.log(0.4)], dtype=np.float32)
cases = [
{
"input": [data1, label1],
"output": expect1,
},
{
"input": [data2, label2],
"output": expect2,
},
]
opr_test(cases, F.cross_entropy)
def test_eye():
dtype = np.float32
cases = [{"input": [10, 20]}, {"input": [20, 30]}]
opr_test(cases,
F.eye,
ref_fn=lambda n, m: np.eye(n, m).astype(dtype),
dtype=dtype)
def test_linspace():
cases = [
{
"input": [1, 9, 9]
},
{
"input": [3, 10, 8]
},
]
opr_test(
cases,
F.linspace,
ref_fn=lambda start, end, step: np.linspace(
start, end, step, dtype=np.float32),
)
cases = [
{
"input": [9, 1, 9]
},
{
"input": [10, 3, 8]
},
]
opr_test(
cases,
F.linspace,
ref_fn=lambda start, end, step: np.linspace(
start, end, step, dtype=np.float32),
)
def common_test_reduce(opr, ref_opr):
data1_shape = (5, 6, 7)
data2_shape = (2, 9, 12)
data1 = np.random.random(data1_shape).astype(np.float32)
data2 = np.random.random(data2_shape).astype(np.float32)
cases = [{"input": data1}, {"input": data2}]
if opr not in (F.argmin, F.argmax):
opr_test(cases, opr, ref_fn=ref_opr)
axis = 2
opr_test(cases, opr, ref_fn=lambda x: ref_opr(x, axis=axis), axis=axis)
axis = 2
keepdims = True
opr_test(
cases,
opr,
ref_fn=lambda x: ref_opr(x, axis=axis, keepdims=keepdims),
axis=axis,
keepdims=keepdims,
)
else:
opr_test(cases, opr, ref_fn=lambda x: ref_opr(x).astype(np.int32))
axis = 2
opr_test(
cases,
opr,
ref_fn=lambda x: ref_opr(x, axis=axis).astype(np.int32),
axis=axis,
)
def test_round():
data1_shape = (15, )
data2_shape = (25, )
data1 = np.random.random(data1_shape).astype(np.float32)
data2 = np.random.random(data2_shape).astype(np.float32)
cases = [{"input": data1}, {"input": data2}]
opr_test(cases, F.round, ref_fn=np.round)
def test_sqrt():
d1_shape = (15, )
d2_shape = (25, )
d1 = np.random.random(d1_shape).astype(np.float32)
d2 = np.random.random(d2_shape).astype(np.float32)
cases = [{"input": d1}, {"input": d2}]
opr_test(cases, F.sqrt, ref_fn=np.sqrt)
def test_matrix_mul():
shape1 = (2, 3)
shape2 = (3, 4)
shape3 = (4, 5)
data1 = np.random.random(shape1).astype("float32")
data2 = np.random.random(shape2).astype("float32")
data3 = np.random.random(shape3).astype("float32")
cases = [{"input": [data1, data2]}, {"input": [data2, data3]}]
opr_test(cases, F.matrix_mul, ref_fn=np.matmul)
def test_logsumexp():
x = np.arange(10).astype(np.float32)
expected = np.log(np.sum(np.exp(x)))
cases = [{"input": x, "output": expected}]
compare_fn = partial(assertTensorClose, allow_special_values=True)
# large value check
n = 100
x = np.full(n, 10000, dtype=np.float32)
expected = 10000 + np.log(n)
cases.append({"input": x, "output": expected.astype(np.float32)})
opr_test(cases, F.logsumexp, axis=0, compare_fn=compare_fn)
# special value check
x = np.array([np.inf], dtype=np.float32)
expected = x
cases = [{"input": x, "output": expected}]
x = np.array([-np.inf, 0.0], dtype=np.float32)
expected = np.zeros(1).astype(np.float32)
cases.append({"input": x, "output": expected})
opr_test(cases, F.logsumexp, axis=0, compare_fn=compare_fn)
x = np.array([np.nan], dtype=np.float32)
expected = x
cases = [{"input": x, "output": expected}]
x = np.array([-np.inf, 1], dtype=np.float32)
expected = np.array([1.0], dtype=np.float32)
cases.append({"input": x, "output": expected})
opr_test(cases, F.logsumexp, axis=0, compare_fn=compare_fn)
# keepdims check
x = np.array([[1e10, 1e-10], [-1e10, -np.inf]], dtype=np.float32)
expected = np.array([[1e10], [-1e10]], dtype=np.float32)
cases = [{"input": x, "output": expected}]
x = np.array([[1e10, -1e-10, 1e-10], [1e10, 1e-10, np.inf]],
dtype=np.float32)
expected = np.array([[1e10], [np.inf]], dtype=np.float32)
cases.append({"input": x, "output": expected})
opr_test(cases, F.logsumexp, axis=1, keepdims=True, compare_fn=compare_fn)
# multiple axes check
x = np.array([[1e10, 1e-10], [-1e10, -np.inf]], dtype=np.float32)
expected = np.array([1e10], dtype=np.float32)
cases = [{"input": x, "output": expected}]
x = np.array([[1e10, -1e-10, 1e-10], [1e10, 1e-10, np.inf]],
dtype=np.float32)
expected = np.array([np.inf], dtype=np.float32)
cases.append({"input": x, "output": expected})
opr_test(cases,
F.logsumexp,
axis=(0, 1),
keepdims=False,
compare_fn=compare_fn)
def test_smooth_l1_loss():
np.random.seed(123)
cases = []
for shape in [(2, 2), (2, 3)]:
data = np.random.uniform(size=shape).astype(np.float32)
label = np.random.uniform(size=shape).astype(np.float32)
diff = np.abs(data - label)
expect = np.where(diff < 1, 0.5 * diff**2, diff - 0.5).mean()
cases.append({"input": [data, label], "output": tensor(expect)})
opr_test(cases, F.smooth_l1_loss)
def test_where():
maskv0 = np.array([[1, 0], [0, 1]], dtype=np.int32)
xv0 = np.array([[1, np.inf], [np.nan, 4]], dtype=np.float32)
yv0 = np.array([[5, 6], [7, 8]], dtype=np.float32)
maskv1 = np.array([[1, 0, 1], [1, 0, 0], [1, 1, 0]], dtype=np.int32)
xv1 = np.array([[1, np.inf, 2], [0, np.nan, 4], [1, 5, 7]],
dtype=np.float32)
yv1 = np.array([[5, 6, 9], [2, 7, 8], [2, 1, 9]], dtype=np.float32)
cases = [{"input": [maskv0, xv0, yv0]}, {"input": [maskv1, xv1, yv1]}]
opr_test(cases, F.where, ref_fn=np.where)
def test_concat():
def get_data_shape(length: int):
return (length, 2, 3)
data1 = np.random.random(get_data_shape(5)).astype("float32")
data2 = np.random.random(get_data_shape(6)).astype("float32")
data3 = np.random.random(get_data_shape(7)).astype("float32")
def run(data1, data2):
return F.concat([data1, data2])
cases = [{"input": [data1, data2]}, {"input": [data1, data3]}]
opr_test(cases, run, ref_fn=lambda x, y: np.concatenate([x, y]))
def test_normalize():
cases = [{
"input": np.random.random((2, 3, 12, 12)).astype(np.float32)
} for i in range(2)]
def np_normalize(x, p=2, axis=None, eps=1e-12):
if axis is None:
norm = np.sum(x**p)**(1.0 / p)
else:
norm = np.sum(x**p, axis=axis, keepdims=True)**(1.0 / p)
return x / np.clip(norm, a_min=eps, a_max=np.inf)
# Test L-2 norm along all dimensions
opr_test(cases, F.normalize, ref_fn=np_normalize)
# Test L-1 norm along all dimensions
opr_test(cases,
partial(F.normalize, p=1),
ref_fn=partial(np_normalize, p=1))
# Test L-2 norm along the second dimension
opr_test(cases,
partial(F.normalize, axis=1),
ref_fn=partial(np_normalize, axis=1))
# Test some norm == 0
cases[0]["input"][0, 0, 0, :] = 0
cases[1]["input"][0, 0, 0, :] = 0
opr_test(cases,
partial(F.normalize, axis=3),
ref_fn=partial(np_normalize, axis=3))
def test_sort():
data1_shape = (10, 3)
data2_shape = (12, 2)
data1 = np.random.random(data1_shape).astype(np.float32)
data2 = np.random.random(data2_shape).astype(np.float32)
output0 = [np.sort(data1), np.argsort(data1).astype(np.int32)]
output1 = [np.sort(data2), np.argsort(data2).astype(np.int32)]
cases = [
{
"input": data1,
"output": output0
},
{
"input": data2,
"output": output1
},
]
opr_test(cases, F.sort)
def test_flatten():
data0_shape = (2, 3, 4, 5)
data1_shape = (4, 5, 6, 7)
data0 = np.random.random(data0_shape).astype(np.float32)
data1 = np.random.random(data1_shape).astype(np.float32)
def compare_fn(x, y):
assert x.numpy().shape == y
output0 = (2 * 3 * 4 * 5, )
output1 = (4 * 5 * 6 * 7, )
cases = [{
"input": data0,
"output": output0
}, {
"input": data1,
"output": output1
}]
opr_test(cases, F.flatten, compare_fn=compare_fn)
output0 = (2, 3 * 4 * 5)
output1 = (4, 5 * 6 * 7)
cases = [{
"input": data0,
"output": output0
}, {
"input": data1,
"output": output1
}]
opr_test(cases, F.flatten, compare_fn=compare_fn, start_axis=1)
output0 = (2, 3, 4 * 5)
output1 = (4, 5, 6 * 7)
cases = [{
"input": data0,
"output": output0
}, {
"input": data1,
"output": output1
}]
opr_test(cases, F.flatten, compare_fn=compare_fn, start_axis=2)
output0 = (2, 3 * 4, 5)
output1 = (4, 5 * 6, 7)
cases = [{
"input": data0,
"output": output0
}, {
"input": data1,
"output": output1
}]
opr_test(cases, F.flatten, compare_fn=compare_fn, start_axis=1, end_axis=2)
def test_batched_matrix_mul():
batch_size = 10
shape1 = (batch_size, 2, 3)
shape2 = (batch_size, 3, 4)
shape3 = (batch_size, 4, 5)
data1 = np.random.random(shape1).astype("float32")
data2 = np.random.random(shape2).astype("float32")
data3 = np.random.random(shape3).astype("float32")
cases = [{"input": [data1, data2]}, {"input": [data2, data3]}]
for i in range(0, batch_size):
def compare_fn(x, y):
x.numpy()[i, ...] == y
opr_test(
cases,
F.batched_matrix_mul,
compare_fn=compare_fn,
ref_fn=lambda x, y: np.matmul(x[i, ...], y[i, ...]),
)
def test_broadcast_to():
input1_shape = (20, 30)
output1_shape = (30, 20, 30)
data1 = np.random.random(input1_shape).astype(np.float32)
input2_shape = (10, 20)
output2_shape = (20, 10, 20)
data2 = np.random.random(input2_shape).astype(np.float32)
def compare_fn(x, y):
assert x.numpy().shape == y
cases = [
{
"input": [data1, output1_shape],
"output": output1_shape
},
{
"input": [data2, output2_shape],
"output": output2_shape
},
]
opr_test(cases, F.broadcast_to, compare_fn=compare_fn)
def common_test_reduce(opr, ref_opr):
data1_shape = (5, 6, 7)
data2_shape = (2, 9, 12)
data1 = np.random.random(data1_shape).astype(np.float32)
data2 = np.random.random(data2_shape).astype(np.float32)
cases = [{"input": data1}, {"input": data2}]
if opr not in (F.argmin, F.argmax):
# test default axis
opr_test(cases, opr, ref_fn=ref_opr)
# test all axises in range of input shape
for axis in range(-3, 3):
# test keepdims False
opr_test(cases,
opr,
ref_fn=lambda x: ref_opr(x, axis=axis),
axis=axis)
# test keepdims True
opr_test(
cases,
opr,
ref_fn=lambda x: ref_opr(x, axis=axis, keepdims=True),
axis=axis,
keepdims=True,
)
else:
# test defaut axis
opr_test(cases, opr, ref_fn=lambda x: ref_opr(x).astype(np.int32))
# test all axises in range of input shape
for axis in range(0, 3):
opr_test(
cases,
opr,
ref_fn=lambda x: ref_opr(x, axis=axis).astype(np.int32),
axis=axis,
)
def test_hinge_loss():
np.random.seed(123)
# case with L1 norm
cases = []
for shape in [(2, 2), (2, 3)]:
data = np.random.uniform(size=shape).astype(np.float32)
label = 2 * np.random.randint(0, 1, size=shape).astype(np.int32) - 1
expect = np.clip(0, np.inf, 1 - data * label).sum(axis=1).mean()
cases.append({"input": [data, label], "output": tensor(expect)})
opr_test(cases, F.hinge_loss)
# cases with L2 norm
cases = []
for shape in [(2, 2), (2, 3)]:
data = np.random.uniform(size=shape).astype(np.float32)
label = 2 * np.random.randint(0, 1, size=shape).astype(np.int32) - 1
expect = ((np.clip(0, np.inf, 1 - data * label)**2).sum(axis=1)).mean()
cases.append({"input": [data, label], "output": tensor(expect)})
def hinge_loss_with_l2_norm(pred, label):
return F.hinge_loss(pred, label, "L2")
opr_test(cases, hinge_loss_with_l2_norm)
def test_arange():
cases = [
{
"input": [1, 9, 1]
},
{
"input": [2, 10, 2]
},
]
opr_test(
cases,
F.arange,
ref_fn=lambda start, end, step: np.arange(
start, end, step, dtype=np.float32),
)
cases = [
{
"input": [9, 1, -1]
},
{
"input": [10, 2, -2]
},
]
opr_test(
cases,
F.arange,
ref_fn=lambda start, end, step: np.arange(
start, end, step, dtype=np.float32),
)
cases = [
{
"input": [9.3, 1.2, -0.5]
},
{
"input": [10.3, 2.1, -1.7]
},
]
opr_test(
cases,
F.arange,
ref_fn=lambda start, end, step: np.arange(
start, end, step, dtype=np.float32),
)
