def ibm_compatible_floats(draw, min_value=None, max_value=None):
if min_value is None:
min_value = MIN_IBM_FLOAT
if max_value is None:
max_value = MAX_IBM_FLOAT
truncated_min_f = max(min_value, MIN_IBM_FLOAT)
truncated_max_f = min(max_value, MAX_IBM_FLOAT)
strategies = []
if truncated_min_f <= LARGEST_NEGATIVE_NORMAL_IBM_FLOAT <= truncated_max_f:
strategies.append(floats(truncated_min_f, LARGEST_NEGATIVE_NORMAL_IBM_FLOAT))
if truncated_min_f <= SMALLEST_POSITIVE_NORMAL_IBM_FLOAT <= truncated_max_f:
strategies.append(floats(SMALLEST_POSITIVE_NORMAL_IBM_FLOAT, truncated_max_f))
if truncated_min_f <= 0 <= truncated_max_f:
strategies.append(just(0.0))
if len(strategies) == 0:
strategies.append(floats(truncated_min_f, truncated_max_f))
ibm = draw(one_of(*strategies))
return ibm
def test_2d_compare_with_numpy(size, nx, xmin, xmax, ny, ymin, ymax):
if xmax <= xmin or ymax <= ymin:
return
x = arrays(np.float, size, elements=st.floats(-1000, 1000)).example()
y = arrays(np.float, size, elements=st.floats(-1000, 1000)).example()
try:
reference = np.histogram2d(x, y, bins=(nx, ny),
range=((xmin, xmax), (ymin, ymax)))[0]
except:
# If Numpy fails, we skip the comparison since this isn't our fault
return
# First, check the Numpy result because it sometimes doesn't make sense. See
# bug report https://github.com/numpy/numpy/issues/9435
n_inside = np.sum((x <= xmax) & (x >= xmin) & (y <= ymax) & (y >= ymin))
if n_inside != np.sum(reference):
return
fast = histogram2d(x, y, bins=(nx, ny),
range=((xmin, xmax), (ymin, ymax)))
print(x, y, nx, xmin, xmax, ny, ymin, ymax)
np.testing.assert_equal(fast, reference)
def _vector_simple_float(self, func, type, data):
func = always_inline(func)
size = rffi.sizeof(rffi.DOUBLE)
myjitdriver = JitDriver(greens = [], reds = 'auto', vectorize=True)
def f(bytecount, va, vb, vc):
i = 0
while i < bytecount:
myjitdriver.jit_merge_point()
a = raw_storage_getitem(type,va,i)
b = raw_storage_getitem(type,vb,i)
c = func(a,b)
raw_storage_setitem(vc, i, rffi.cast(type,c))
i += size
la = data.draw(st.lists(st.floats(), min_size=10, max_size=150))
l = len(la)
lb = data.draw(st.lists(st.floats(), min_size=l, max_size=l))
rawstorage = RawStorage()
va = rawstorage.new(la, type)
vb = rawstorage.new(lb, type)
vc = rawstorage.new(None, type, size=l)
self.meta_interp(f, [l*size, va, vb, vc], vec=True)
for i in range(l):
c = raw_storage_getitem(type,vc,i*size)
r = rffi.cast(type, func(la[i], lb[i]))
assert isclose(r, c)
rawstorage.clear()
def _inputs(draw):
N = draw(st.integers(min_value=0, max_value=5))
D = draw(st.integers(min_value=1, max_value=5))
# N, D, data, lambda1, lambda2
return (
N,
D,
draw(st.lists(
min_size=N * D,
max_size=N * D,
elements=st.one_of(
st.floats(min_value=-10, max_value=1 - TOLERANCE),
st.floats(min_value=1 + TOLERANCE, max_value=10))
)),
draw(st.lists(
elements=st.one_of(
st.floats(min_value=-2, max_value=-TOLERANCE),
st.floats(min_value=TOLERANCE, max_value=2)),
min_size=D,
max_size=D,
)),
draw(st.lists(
elements=st.floats(min_value=-2, max_value=2),
min_size=D,
max_size=D,
)),
)
def test_power_measurement(instrument, data):
channel_id = data.draw(sampled_from(instrument.channel_ids))
channel = instrument.channel(channel_id)
voltage = data.draw(floats(channel.voltage.protection.min, channel.voltage.protection.max).map(lambda v: round(v, 3)))
current = data.draw(floats(channel.current.protection.min, channel.current.protection.max).map(lambda v: round(v, 3)))
instrument._inst._channel_voltage_measurements[channel_id] = voltage
instrument._inst._channel_current_measurements[channel_id] = current
assert channel.power.measurement == round(voltage*current, 3)
def test_out_of_range_floats_are_bad():
with pytest.raises(BadData):
floats(11, 12).from_basic(floats(0, 1).to_basic((0, 0.0)))
with pytest.raises(BadData):
FixedBoundedFloatStrategy(11, 12).from_basic(
floats().to_basic(float(u'nan'))
)
def test_diff_values_float(data):
x = data.draw(st.just(0), label='x')
y = data.draw(st.floats(min_value=1e8), label='y')
z = data.draw(st.floats(max_value=-1e8), label='z')
assert not are_values_different(x, x)
assert are_values_different(x, y)
assert are_values_different(x, z)
assert are_values_different(y, z)
def dataset_utm_north_down(draw):
"""Generate a fake UTM dataset with an origin, a resolution, and a finite size"""
x = draw(floats(min_value=-1e6, max_value=1e+6, allow_nan=False, allow_infinity=False))
y = draw(floats(min_value=-1e6, max_value=1e+6, allow_nan=False, allow_infinity=False))
res = draw(floats(min_value=0.1, max_value=30, allow_nan=False, allow_infinity=False))
h = draw(integers(min_value=1, max_value=1000))
w = draw(integers(min_value=1, max_value=1000))
return FakeDataset(
transform=windows.Affine.identity() * windows.Affine.translation(x, y) * windows.Affine.scale(res),
height=h, width=w)
def test_no_single_floats_in_range():
low = 2.0 ** 25 + 1
high = low + 2
st.floats(low, high).validate() # Note: OK for 64bit floats
with pytest.raises(InvalidArgument):
"""Unrepresentable bounds are deprecated; but we're not testing that
here."""
with warnings.catch_warnings():
warnings.simplefilter("ignore")
st.floats(low, high, width=32).validate()
def test_diff_values_array(data):
a = data.draw(st.lists(elements=st.integers(min_value=0), min_size=1))
b = data.draw(st.lists(elements=st.integers(max_value=-1), min_size=1))
c = data.draw(st.lists(elements=st.floats(min_value=1e8), min_size=1))
d = data.draw(st.lists(elements=st.floats(max_value=-1e8), min_size=1))
# TODO: Figure out a way to include 0 in lists (arrays)
assert are_values_different(a, b)
assert are_values_different(c, d)
assert not are_values_different(a, a)
def from_dtype(dtype):
# type: (np.dtype) -> st.SearchStrategy[Any]
"""Creates a strategy which can generate any value of the given dtype."""
check_type(np.dtype, dtype, 'dtype')
# Compound datatypes, eg 'f4,f4,f4'
if dtype.names is not None:
# mapping np.void.type over a strategy is nonsense, so return now.
return st.tuples(
*[from_dtype(dtype.fields[name][0]) for name in dtype.names])
# Subarray datatypes, eg '(2, 3)i4'
if dtype.subdtype is not None:
subtype, shape = dtype.subdtype
return arrays(subtype, shape)
# Scalar datatypes
if dtype.kind == u'b':
result = st.booleans() # type: SearchStrategy[Any]
elif dtype.kind == u'f':
if dtype.itemsize == 2:
result = st.floats(width=16)
elif dtype.itemsize == 4:
result = st.floats(width=32)
else:
result = st.floats()
elif dtype.kind == u'c':
if dtype.itemsize == 8:
float32 = st.floats(width=32)
result = st.builds(complex, float32, float32)
else:
result = st.complex_numbers()
elif dtype.kind in (u'S', u'a'):
# Numpy strings are null-terminated; only allow round-trippable values.
# `itemsize == 0` means 'fixed length determined at array creation'
result = st.binary(max_size=dtype.itemsize or None
).filter(lambda b: b[-1:] != b'\0')
elif dtype.kind == u'u':
result = st.integers(min_value=0,
max_value=2 ** (8 * dtype.itemsize) - 1)
elif dtype.kind == u'i':
overflow = 2 ** (8 * dtype.itemsize - 1)
result = st.integers(min_value=-overflow, max_value=overflow - 1)
elif dtype.kind == u'U':
# Encoded in UTF-32 (four bytes/codepoint) and null-terminated
result = st.text(max_size=(dtype.itemsize or 0) // 4 or None
).filter(lambda b: b[-1:] != u'\0')
elif dtype.kind in (u'm', u'M'):
if '[' in dtype.str:
res = st.just(dtype.str.split('[')[-1][:-1])
else:
res = st.sampled_from(TIME_RESOLUTIONS)
result = st.builds(dtype.type, st.integers(-2**63, 2**63 - 1), res)
else:
raise InvalidArgument(u'No strategy inference for {}'.format(dtype))
return result.map(dtype.type)
def prob_end_spatial_tournaments(draw, strategies=strategies,
min_size=1, max_size=10,
min_prob_end=0, max_prob_end=1,
min_noise=0, max_noise=1,
min_repetitions=1, max_repetitions=20):
"""
A hypothesis decorator to return a probabilistic ending spatial tournament.
Parameters
----------
min_size : integer
The minimum number of strategies to include
max_size : integer
The maximum number of strategies to include
min_prob_end : float
The minimum probability of a match ending
max_prob_end : float
The maximum probability of a match ending
min_noise : float
The minimum noise value
max_noise : float
The maximum noise value
min_repetitions : integer
The minimum number of repetitions
max_repetitions : integer
The maximum number of repetitions
"""
strategies = draw(strategy_lists(strategies=strategies,
min_size=min_size,
max_size=max_size))
players = [s() for s in strategies]
player_indices = list(range(len(players)))
all_potential_edges = list(itertools.combinations(player_indices, 2))
all_potential_edges.extend([(i, i) for i in player_indices]) # Loops
edges = draw(lists(sampled_from(all_potential_edges), unique=True,
average_size=2 * len(players)))
# Ensure all players/nodes are connected:
node_indices = sorted(set([node for edge in edges for node in edge]))
missing_nodes = [index
for index in player_indices if index not in node_indices]
for index in missing_nodes:
opponent = draw(sampled_from(player_indices))
edges.append((index, opponent))
prob_end = draw(floats(min_value=min_prob_end, max_value=max_prob_end))
repetitions = draw(integers(min_value=min_repetitions,
max_value=max_repetitions))
noise = draw(floats(min_value=min_noise, max_value=max_noise))
tournament = ProbEndSpatialTournament(players, prob_end=prob_end,
repetitions=repetitions,
noise=noise, edges=edges)
return tournament
def valid_phase_type_generator(draw, k):
ptg = []
for i in range(k):
ptg.append(draw(lists(floats(min_value=0, max_value=1),
min_size=k, max_size=k)))
ptg_array = np.array(ptg)
ptg_totals = np.array(draw(lists(floats(min_value = 0, max_value=1),
min_size=k, max_size=k))).reshape(-1, 1)
with np.errstate(invalid='ignore'):
ptg_norm = (np.nan_to_num(ptg_array / ptg_array.sum(axis=1).reshape(-1,1))) \
* ptg_totals
return ptg_norm
def sample_row(draw):
"""Get a sample database row."""
return OrderedDict((
('id', draw(st.integers(min_value=1))),
('lat', draw(st.floats(min_value=-180, max_value=180))),
('lon', draw(st.floats(min_value=-90, max_value=90))),
('country', draw(st.sampled_from(['Cintra', 'Arnor', 'Arrakis', 'Gondor']))),
('town', draw(st.sampled_from(['Caer Morhen', 'Minas Tirith', 'Sietch Tabr', 'Gondolin']))),
('postcode', draw(st.sampled_from(['123123', '23423', '123122', '43223', '231232']))),
('street', draw(st.sampled_from(['street1', 'street2', 'street3', 'street4', 'street5']))),
('house', draw(st.integers(max_value=500))),
))
def markov_chains(draw, alphabets=((2, 4), (2, 4), (2, 4))):
"""
Generate Markov chains for use with hypothesis.
Parameters
----------
draw : function
A sampling function passed in by hypothesis.
alphabets : int, tuple of ints, tuple of pairs of ints
If an int, it is the length of the chain and each variable is assumed to be binary.
If a tuple of ints, the ints are assumed to be the size of each variable. If a tuple
of pairs of ints, each pair represents the min and max alphabet size of each variable.
Returns
-------
dist : Distribution
A Markov chain with variable sizes.
"""
try:
len(alphabets)
try:
len(alphabets[0])
except TypeError:
alphabets = tuple((alpha, alpha) for alpha in alphabets)
except TypeError:
alphabets = ((2, 2),)*alphabets
alphabets = [int(draw(integers(*alpha))) for alpha in alphabets]
px = draw(arrays(np.float, shape=alphabets[0], elements=floats(0, 1)))
cds = [draw(arrays(np.float, shape=(a, b), elements=floats(0, 1))) for a, b in pairwise(alphabets)]
# assume things
assume(px.sum() > 0)
for cd in cds:
for row in cd:
assume(row.sum() > 0)
px /= px.sum()
# construct dist
for cd in cds:
cd /= cd.sum(axis=1, keepdims=True)
slc = (np.newaxis,)*(len(px.shape)-1) + (colon, colon)
px = px[..., np.newaxis] * cd[slc]
dist = Distribution.from_ndarray(px)
dist.normalize()
return dist
def float_range(self, left, right):
for f in (math.isnan, math.isinf):
for x in (left, right):
assume(not f(x))
left, right = sorted((left, right))
assert left <= right
return strategy(floats(left, right))
def _gen_df(n_rows=1000):
"""Generate a random DataFrame.
.. note::
Generating a random value for every row takes too long,
so we use the same value for every row.
This may not be the safest way to do it though!
"""
logger.debug("_gen_df(%s)", n_rows)
_get_finite_value = functools.partial(_get_numeric_value, filt=np.isfinite)
data = {
'integer': np.repeat(_get_finite_value(
st.integers(
min_value=np.iinfo(np.int64).min * 0.99,
max_value=np.iinfo(np.int64).max * 0.99)),
n_rows),
'float': np.repeat(_get_finite_value(
st.floats(
min_value=np.finfo(np.float64).min * 0.99,
max_value=np.finfo(np.float64).max * 0.99)),
n_rows),
'bool': np.repeat(_get_finite_value(st.booleans()), n_rows),
'text': np.repeat(
st.text(alphabet=string.ascii_letters, min_size=0, max_size=100).example(),
n_rows),
}
df = pd.DataFrame(data)
# print(df.head())
return df
def host_json():
return st.fixed_dictionaries(
{
"metadata":
st.fixed_dictionaries({
"update_time": st.floats(),
"update_user": st.one_of(st.none(), st.text()),
"update_action": st.integers(),
"creator": st.text(),
"create_time": st.integers(),
"update_controller_action": st.text(),
"owner": st.one_of(st.none(), st.text()),
"command_id": st.one_of(st.none(), st.text(), st.integers()),}),
"name": st.one_of(st.none(), st.text()),
"ip": st.one_of(st.none(), st.text()),
"_rev": st.one_of(st.none(), st.text()),
"description": st.one_of(st.none(), st.text()),
"default_gateway": st.one_of(st.none(), st.text()),
"owned": st.booleans(),
"services": st.one_of(st.none(), st.integers()),
"hostnames": st.lists(st.text()),
"vulns": st.one_of(st.none(), st.integers()),
"owner": st.one_of(st.none(), st.text()),
"credentials": st.one_of(st.none(), st.integers()),
"_id": st.one_of(st.none(), st.integers()),
"os": st.one_of(st.none(), st.text()),
"id": st.one_of(st.none(), st.integers()),
"icon": st.one_of(st.none(), st.text())}
)
def test_flatmap_retrieve_from_db():
constant_float_lists = strategy(floats(0, 1)).flatmap(
lambda x: lists(just(x))
)
track = []
db = ExampleDatabase()
@given(constant_float_lists, settings=Settings(database=db))
def record_and_test_size(xs):
track.append(xs)
assert sum(xs) < 1
with pytest.raises(AssertionError):
record_and_test_size()
assert track
example = track[-1]
while track:
track.pop()
with pytest.raises(AssertionError):
record_and_test_size()
assert track[0] == example
def python_number(draw, min_val, max_val):
return draw(st.one_of(st.floats(min_val,
max_val,
allow_nan=False,
allow_infinity=False),
st.integers(min_val,
max_val)))
def _dense_features_map(draw, num_records, **kwargs):
float_lengths = draw(
st.lists(
st.integers(min_value=1, max_value=10),
min_size=num_records,
max_size=num_records
)
)
total_length = sum(float_lengths)
float_keys = draw(
st.lists(
st.integers(min_value=1, max_value=100),
min_size=total_length,
max_size=total_length,
unique=True
)
)
float_values = draw(
st.lists(st.floats(),
min_size=total_length,
max_size=total_length)
)
return [float_lengths, float_keys, float_values]
def test_floats_in_tiny_interval_within_bounds(data, center):
assume(not (math.isinf(next_down(center)) or math.isinf(next_up(center))))
lo = Decimal.from_float(next_down(center)).next_plus()
hi = Decimal.from_float(next_up(center)).next_minus()
assert float(lo) < lo < center < hi < float(hi)
val = data.draw(st.floats(lo, hi))
assert lo < val < hi
def matches(draw, strategies=axelrod.strategies,
min_turns=1, max_turns=200,
min_noise=0, max_noise=1):
"""
A hypothesis decorator to return a random match as well as a random seed (to
ensure reproducibility when instance of class need the random library).
Parameters
----------
strategies : list
The strategies from which to sample the two the players
min_turns : integer
The minimum number of turns
max_turns : integer
The maximum number of turns
min_noise : float
The minimum noise
max_noise : float
The maximum noise
Returns
-------
tuple : a random match as well as a random seed
"""
seed = draw(random_module())
strategies = draw(strategy_lists(min_size=2, max_size=2))
players = [s() for s in strategies]
turns = draw(integers(min_value=min_turns, max_value=max_turns))
noise = draw(floats(min_value=min_noise, max_value=max_noise))
match = axelrod.Match(players, turns=turns, noise=noise)
return match, seed
def test_range_check_returns_range_as_is_if_first_is_less_than_second(x, d):
# Pull data such that the first is less than the second.
if isinstance(x, float):
y = d.draw(floats(min_value=x + 1.0, max_value=1E9, allow_nan=False))
else:
y = d.draw(integers(min_value=x + 1))
assert range_check(x, y) == (x, y)
def matches(
draw,
strategies=short_run_time_strategies,
min_turns=1,
max_turns=200,
min_noise=0,
max_noise=1,
):
"""
A hypothesis decorator to return a random match.
Parameters
----------
strategies : list
The strategies from which to sample the two the players
min_turns : integer
The minimum number of turns
max_turns : integer
The maximum number of turns
min_noise : float
The minimum noise
max_noise : float
The maximum noise
Returns
-------
match : a random match
"""
strategies = draw(strategy_lists(min_size=2, max_size=2))
players = [s() for s in strategies]
turns = draw(integers(min_value=min_turns, max_value=max_turns))
noise = draw(floats(min_value=min_noise, max_value=max_noise))
match = Match(players, turns=turns, noise=noise)
return match
def test_small_sum_lists():
xs = minimal(
lists(floats()),
lambda x: len(x) >= 100 and sum(t for t in x if float("inf") > t >= 0) >= 1,
settings=Settings(average_list_length=200),
)
assert 1.0 <= sum(t for t in xs if t >= 0) <= 1.5
def test_may_fill_with_nan_when_unique_is_set():
find_any(
nps.arrays(
dtype=float, elements=st.floats(allow_nan=False), shape=10,
unique=True, fill=st.just(float('nan'))),
lambda x: np.isnan(x).any()
)
def _tensor_splits(draw):
lengths = draw(st.lists(st.integers(1, 5), min_size=1, max_size=10))
batch_size = draw(st.integers(1, 5))
element_pairs = [
(batch, r) for batch in range(batch_size) for r in range(len(lengths))
]
perm = draw(st.permutations(element_pairs))
perm = perm[:-1] # skip one range
ranges = [[(0, 0)] * len(lengths) for _ in range(batch_size)]
offset = 0
for pair in perm:
ranges[pair[0]][pair[1]] = (offset, lengths[pair[1]])
offset += lengths[pair[1]]
data = draw(st.lists(
st.floats(min_value=-1.0, max_value=1.0),
min_size=offset,
max_size=offset
))
key = draw(st.permutations(range(offset)))
return (
np.array(data).astype(np.float32), np.array(ranges),
np.array(lengths), np.array(key).astype(np.int64)
)
def test_flatmap_retrieve_from_db():
constant_float_lists = floats(0, 1).flatmap(
lambda x: lists(just(x))
)
track = []
db = ExampleDatabase()
@given(constant_float_lists)
@settings(database=db)
def record_and_test_size(xs):
if sum(xs) >= 1:
track.append(xs)
assert False
with pytest.raises(AssertionError):
record_and_test_size()
assert track
example = track[-1]
track = []
with pytest.raises(AssertionError):
record_and_test_size()
assert track[0] == example
def reusable():
return st.one_of(
st.sampled_from(base_reusable_strategies),
st.builds(
st.floats, min_value=st.none() | st.floats(),
max_value=st.none() | st.floats(), allow_infinity=st.booleans(),
allow_nan=st.booleans()
),
st.builds(st.just, st.lists(max_size=0)),
st.builds(st.sampled_from, st.lists(st.lists(max_size=0))),
st.lists(reusable).map(st.one_of),
st.lists(reusable).map(lambda ls: st.tuples(*ls)),
)
class TestCrossEntropyOps(hu.HypothesisTestCase):
@given(
inputs=st.lists(
elements=st.integers(min_value=1, max_value=5),
min_size=1,
max_size=2,
average_size=2,
).flatmap(
lambda shape: st.tuples(
hu.arrays(
dims=shape,
elements=st.one_of(
st.floats(min_value=-1.0, max_value=-0.1),
st.floats(min_value=0.1, max_value=1.0),
)),
hu.arrays(
dims=shape,
elements=st.sampled_from([0.0, 1.0]),
),
)
),
)
def test_sigmoid_cross_entropy_with_logits(self, inputs):
logits, targets = inputs
def sigmoid_xentr_logit_ref(logits, targets):
s = sigmoid_cross_entropy_with_logits(logits, targets)
m = np.mean(s, axis=len(logits.shape) - 1)
return (m, )
def sigmoid_xentr_logit_grad_ref(g_out, outputs, fwd_inputs):
fwd_logits, fwd_targets = fwd_inputs
inner_size = fwd_logits.shape[-1]
m = fwd_targets - sigmoid(fwd_logits)
g_in = -np.expand_dims(g_out, axis=-1) * m / inner_size
return (g_in, None)
op = core.CreateOperator(
'SigmoidCrossEntropyWithLogits',
['logits', 'targets'],
['xentropy'])
self.assertReferenceChecks(
hu.cpu_do,
op,
[logits, targets],
sigmoid_xentr_logit_ref,
output_to_grad='xentropy',
grad_reference=sigmoid_xentr_logit_grad_ref)
@given(n=st.integers(2, 10),
b=st.integers(1, 5),
**hu.gcs_cpu_only)
def test_soft_label_cross_entropy(self, n, b, gc, dc):
# Initialize X and add 1e-2 for numerical stability
X = np.random.rand(b, n).astype(np.float32)
X = X + 1e-2
for i in range(b):
X[i] = X[i] / np.sum(X[i])
# Initialize label
label = np.random.rand(b, n).astype(np.float32)
for i in range(b):
label[i] = label[i] / np.sum(label[i])
# Reference implementation of cross entropy with soft labels
def soft_label_xentr_ref(X, label):
xent = [np.sum((-label[j][i] * np.log(max(X[j][i], 1e-20))
for i in range(len(X[0])))) for j in range(b)]
return (xent,)
op = core.CreateOperator("CrossEntropy", ["X", "label"], ["Y"])
# TODO(surya) Once CrossEntropyOp is ported to GPU, add the respective
# tests to this unit test.
self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[X, label],
reference=soft_label_xentr_ref,
)
self.assertGradientChecks(
gc, op, [X, label], 0, [0], stepsize=1e-4, threshold=1e-2)
import numpy as np
from hypothesis import assume, given, settings
from hypothesis.strategies import floats
from nxviz.polcart import (
to_cartesian,
to_degrees,
to_polar,
to_proper_degrees,
to_proper_radians,
to_radians,
)
@settings(perform_health_check=False)
@given(floats(), floats())
def test_convert_xy(x, y):
assume(x != 0 and y != 0)
assume(np.isfinite(x) and np.isfinite(y))
assume(abs(x) < 1E6 and abs(y) < 1E6)
assume(abs(x) > 0.01 and abs(y) > 0.01)
# Test radians
r, theta = to_polar(x, y)
x_new, y_new = to_cartesian(r, theta)
assert np.allclose(x, x_new)
assert np.allclose(y, y_new)
# Test degrees
r, theta = to_polar(x, y, theta_units="degrees")
x_new, y_new = to_cartesian(r, theta, theta_units="degrees")
class TestCrossEntropyOps(hu.HypothesisTestCase):
@given(
inputs=st.lists(
elements=st.integers(min_value=1, max_value=5),
min_size=1,
max_size=2,
average_size=2,
).flatmap(
lambda shape: st.tuples(
hu.arrays(
dims=shape,
elements=st.one_of(
st.floats(min_value=-1.0, max_value=-0.1),
st.floats(min_value=0.1, max_value=1.0),
)),
hu.arrays(
dims=shape,
elements=st.sampled_from([0.0, 1.0]),
),
)
),
options=st.one_of(
st.tuples(st.just(True), st.just(False)),
st.tuples(st.just(False), st.just(True)),
st.tuples(st.just(False), st.just(False))
),
**hu.gcs
)
def test_sigmoid_cross_entropy_with_logits(
self, inputs, options, gc, dc
):
logits, targets = inputs
log_D_trick, unjoined_lr_loss = options
def sigmoid_xentr_logit_ref(logits, targets):
if unjoined_lr_loss:
s = unjoined_sigmoid_cross_entropy(logits, targets)
else:
s = (
sigmoid_cross_entropy_with_logits(logits, targets)
if not log_D_trick else
sigmoid_cross_entropy_with_logits_with_log_D_trick(
logits, targets
)
)
m = np.mean(s, axis=len(logits.shape) - 1)
return (m, )
def sigmoid_xentr_logit_grad_ref(g_out, outputs, fwd_inputs):
fwd_logits, fwd_targets = fwd_inputs
inner_size = fwd_logits.shape[-1]
if unjoined_lr_loss:
m = unjoined_sigmoid_cross_entropy_grad(logits, targets)
else:
m = (
sigmoid_cross_entropy_with_logits_grad(fwd_logits, fwd_targets)
if not log_D_trick else
sigmoid_cross_entropy_with_logits_with_log_D_trick_grad(
fwd_logits, fwd_targets
)
)
# m = fwd_targets - sigmoid(fwd_logits)
g_in = -np.expand_dims(g_out, axis=-1) * m / inner_size
return (g_in, None)
op = core.CreateOperator(
'SigmoidCrossEntropyWithLogits', ['logits', 'targets'],
['xentropy'],
log_D_trick=log_D_trick,
unjoined_lr_loss=unjoined_lr_loss
)
self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[logits, targets],
reference=sigmoid_xentr_logit_ref,
output_to_grad='xentropy',
grad_reference=sigmoid_xentr_logit_grad_ref)
@given(
log_D_trick=st.just(False),
**hu.gcs_cpu_only
)
def test_cross_entropy_and_unjoied_cross_entropy_relation(
self, log_D_trick, gc, dc
):
logits = np.array([1.4720, 0.3500, -0.6529, -1.1908, 0.8357,
-1.0774, -0.3395, -0.2469, 0.6708, -1.8332], dtype='f')
targets = np.array([1., 1., 1., 1., 1., 1., 0., 0., 0., 0.], dtype='f')
lr_size = targets.size
unjoined_lr_loss = False
def sigmoid_xentr_logit_ref(logits, targets):
if unjoined_lr_loss:
s = unjoined_sigmoid_cross_entropy(logits, targets)
else:
s = sigmoid_cross_entropy_with_logits(logits, targets)
m = np.mean(s, axis=len(logits.shape) - 1)
return (m, )
def sigmoid_xentr_logit_grad_ref(g_out, outputs, fwd_inputs):
fwd_logits, fwd_targets = fwd_inputs
inner_size = fwd_logits.shape[-1]
if unjoined_lr_loss:
m = unjoined_sigmoid_cross_entropy_grad(logits, targets)
else:
m = sigmoid_cross_entropy_with_logits_grad(
fwd_logits, fwd_targets)
# m = fwd_targets - sigmoid(fwd_logits)
g_in = -np.expand_dims(g_out, axis=-1) * m / inner_size
return (g_in, None)
op = core.CreateOperator(
'SigmoidCrossEntropyWithLogits', ['logits', 'targets'],
['xentropy'],
log_D_trick=log_D_trick,
unjoined_lr_loss=unjoined_lr_loss
)
output_lr = self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[logits, targets],
reference=sigmoid_xentr_logit_ref,
output_to_grad='xentropy',
grad_reference=sigmoid_xentr_logit_grad_ref)
# Unjoined dataset where labels change later
logits = np.array([1.4720, 0.3500, -0.6529, -1.1908, 0.8357,
-1.0774, -0.3395, -0.2469, 0.6708, -1.8332, 1.4720, 0.3500,
-0.6529, -1.1908, 0.8357, -1.0774], dtype='f')
targets = np.array([0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 1., 1., 1., 1., 1., 1.], dtype='f')
unjoined_lr_loss = True
unjoined_lr_size = targets.size
op = core.CreateOperator(
'SigmoidCrossEntropyWithLogits', ['logits', 'targets'],
['xentropy'],
log_D_trick=log_D_trick,
unjoined_lr_loss=unjoined_lr_loss
)
outputs_unjoined_lr = self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[logits, targets],
reference=sigmoid_xentr_logit_ref,
output_to_grad='xentropy',
grad_reference=sigmoid_xentr_logit_grad_ref)
self.assertAlmostEqual(
output_lr[0].item(0) * lr_size / unjoined_lr_size,
outputs_unjoined_lr[0].item(0),
delta=0.0001)
@given(
inputs=st.lists(
elements=st.integers(min_value=1, max_value=5),
min_size=1,
max_size=2,
average_size=2,
).flatmap(
lambda shape: st.tuples(
hu.arrays(
dims=shape,
elements=st.one_of(
st.floats(min_value=-1.0, max_value=-0.1),
st.floats(min_value=0.1, max_value=1.0),
)),
hu.arrays(
dims=shape,
elements=st.sampled_from([0.0, 1.0]),
),
hu.arrays(
dims=shape,
elements=st.floats(min_value=0.1, max_value=1.0),
),
)
),
**hu.gcs
)
def test_weighted_sigmoid_cross_entropy_with_logits(self, inputs, gc, dc):
logits, targets, weights = inputs
def weighted_sigmoid_xentr_logit_ref(logits, targets, weights):
s = sigmoid_cross_entropy_with_logits(logits, targets)
s = np.multiply(s, weights)
m = np.mean(s, axis=len(logits.shape) - 1)
return (m, )
def weighted_sigmoid_xentr_logit_grad_ref(g_out, outputs, fwd_inputs):
fwd_logits, fwd_targets, fwd_weights = fwd_inputs
inner_size = fwd_logits.shape[-1]
m = fwd_targets - sigmoid(fwd_logits)
m = np.multiply(m, weights)
g_in = -np.expand_dims(g_out, axis=-1) * m / inner_size
return (g_in, None, None)
op = core.CreateOperator(
'WeightedSigmoidCrossEntropyWithLogits',
['logits', 'targets', 'weights'],
['xentropy'])
self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[logits, targets, weights],
reference=weighted_sigmoid_xentr_logit_ref,
output_to_grad='xentropy',
grad_reference=weighted_sigmoid_xentr_logit_grad_ref)
@given(n=st.integers(2, 10),
b=st.integers(1, 5),
**hu.gcs_cpu_only)
def test_soft_label_cross_entropy(self, n, b, gc, dc):
# Initialize X and add 1e-2 for numerical stability
X = np.random.rand(b, n).astype(np.float32)
X = X + 1e-2
for i in range(b):
X[i] = X[i] / np.sum(X[i])
# Initialize label
label = np.random.rand(b, n).astype(np.float32)
for i in range(b):
label[i] = label[i] / np.sum(label[i])
# Reference implementation of cross entropy with soft labels
def soft_label_xentr_ref(X, label):
xent = [np.sum((-label[j][i] * np.log(max(X[j][i], 1e-20))
for i in range(len(X[0])))) for j in range(b)]
return (xent,)
op = core.CreateOperator("CrossEntropy", ["X", "label"], ["Y"])
# TODO(surya) Once CrossEntropyOp is ported to GPU, add the respective
# tests to this unit test.
self.assertReferenceChecks(
device_option=gc,
op=op,
inputs=[X, label],
reference=soft_label_xentr_ref,
)
self.assertGradientChecks(
gc, op, [X, label], 0, [0], stepsize=1e-4, threshold=1e-2)
from hypothesis.strategies import just, none, sets, text, basic, lists, \
binary, builds, floats, one_of, tuples, randoms, booleans, decimals, \
integers, composite, fractions, recursive, streaming, frozensets, \
dictionaries, sampled_from, complex_numbers, fixed_dictionaries
from hypothesis.strategytests import mutate_basic, templates_for, \
strategy_test_suite
from hypothesis.internal.compat import hrange, OrderedDict
from hypothesis.searchstrategy.morphers import MorpherStrategy
from hypothesis.searchstrategy.narytree import n_ary_tree
with Settings(average_list_length=5.0):
TestIntegerRange = strategy_test_suite(integers(min_value=0, max_value=5))
TestGiantIntegerRange = strategy_test_suite(
integers(min_value=(-(2 ** 129)), max_value=(2 ** 129))
)
TestFloatRange = strategy_test_suite(floats(min_value=0.5, max_value=10))
TestSampled10 = strategy_test_suite(sampled_from(elements=list(range(10))))
TestSampled1 = strategy_test_suite(sampled_from(elements=(1,)))
TestSampled2 = strategy_test_suite(sampled_from(elements=(1, 2)))
TestIntegersFrom = strategy_test_suite(integers(min_value=13))
TestIntegersFrom = strategy_test_suite(integers(min_value=1 << 1024))
TestOneOf = strategy_test_suite(one_of(
integers(), integers(), booleans()))
TestOneOfSameType = strategy_test_suite(
one_of(
integers(min_value=1, max_value=10),
integers(min_value=8, max_value=15),
)
assert x + y == y + x
@fails
@given(binary(), binary())
def test_bytes_addition_is_commutative(x, y):
assert x + y == y + x
@given(integers(), integers(), integers())
def test_int_addition_is_associative(x, y, z):
assert x + (y + z) == (x + y) + z
@fails
@given(floats(), floats(), floats())
@settings(max_examples=2000,)
def test_float_addition_is_associative(x, y, z):
assert x + (y + z) == (x + y) + z
@given(lists(integers()))
def test_reversing_preserves_integer_addition(xs):
assert sum(xs) == sum(reversed(xs))
def test_still_minimizes_on_non_assertion_failures():
@settings(max_examples=50)
@given(integers())
def is_not_too_large(x):
if x >= 10:
class SimulatorsActAsFactory(RuleBasedStateMachine):
def __init__(self):
super().__init__()
products, profiles = sample_products, sample_profiles
self.slow_simulator = SlowFactorySimulator(
initial_wallet=initial_wallet,
initial_storage=initial_storage,
n_steps=n_steps,
n_products=len(products),
profiles=profiles,
max_storage=max_storage,
)
self.fast_simulator = FastFactorySimulator(
initial_wallet=initial_wallet,
initial_storage=initial_storage,
n_steps=n_steps,
n_products=len(products),
profiles=profiles,
max_storage=max_storage,
)
self.factory = Factory(
initial_wallet=initial_wallet,
initial_storage=initial_storage,
profiles=profiles,
max_storage=max_storage,
)
self.profiles = self.factory.profiles
self._payments = defaultdict(list)
self._transports = defaultdict(list)
profile_indices = Bundle("profile_indices")
payments = Bundle("payments")
times = Bundle("times")
transports = Bundle("transports")
@rule(target=profile_indices, k=st.integers(len(sample_profiles)))
def choose_profile(self, k):
return k
@rule(targe=payments, payment=st.floats(-30, 30))
def choose_payment(self, payment):
return payment
@rule(targe=times, payment=st.integers(0, n_steps))
def choose_time(self, t):
return t
@rule(targe=transports, storage=storage())
def choose_transport(self, t):
return t
@rule(profile_index=profile_indices, t=times)
def run_profile(self, profile_index, t):
job = Job(
profile=profile_index,
time=t,
line=self.profiles[profile_index].line,
action="run",
contract=None,
override=False,
)
end = t + self.profiles[profile_index].n_steps
if end > n_steps - 1:
return
self.factory.schedule(job=job, override=False)
self.slow_simulator.schedule(job=job, override=False)
self.fast_simulator.schedule(job=job, override=False)
@rule(payment=payments, t=times)
def pay(self, payment, t):
self._payments[t].append(payment)
self.slow_simulator.pay(payment, t)
self.fast_simulator.pay(payment, t)
@rule(trans=transports, t=times)
def transport(self, trans, t):
p, q = trans
self._transports[t].append(trans)
self.slow_simulator.transport_to(p, q, t)
self.fast_simulator.transport_to(p, q, t)
@rule()
def run_and_test(self):
for _ in range(n_steps):
for payment in self._payments[_]:
self.factory.pay(payment)
for p, q in self._transports[_]:
self.factory.transport_to(p, q)
self.factory.step()
assert (self.slow_simulator.wallet_at(_) ==
self.fast_simulator.wallet_at(_) == self.factory.wallet)
assert (self.slow_simulator.balance_at(_) ==
self.fast_simulator.balance_at(_) == self.factory.balance)
assert np.all(
self.slow_simulator.storage_at(_) ==
self.fast_simulator.storage_at(_))
assert np.all(
self.slow_simulator.storage_at(_) == storage_as_array(
self.factory.storage, n_products=len(sample_products)))
assert np.all(
self.slow_simulator.line_schedules_at(_) ==
self.factory.line_schedules)
assert np.all(
self.fast_simulator.line_schedules_at(_) ==
self.factory.line_schedules)
def schur_complements(draw,
batch_shape=None,
event_dim=None,
feature_dim=None,
feature_ndims=None,
enable_vars=None,
depth=None):
"""Strategy for drawing `SchurComplement` kernels.
The underlying kernel is drawn from the `kernels` strategy.
Args:
draw: Hypothesis strategy sampler supplied by `@hps.composite`.
batch_shape: An optional `TensorShape`. The batch shape of the resulting
Kernel. Hypothesis will pick a batch shape if omitted.
event_dim: Optional Python int giving the size of each of the
kernel's parameters' event dimensions. This is shared across all
parameters, permitting square event matrices, compatible location and
scale Tensors, etc. If omitted, Hypothesis will choose one.
feature_dim: Optional Python int giving the size of each feature dimension.
If omitted, Hypothesis will choose one.
feature_ndims: Optional Python int stating the number of feature dimensions
inputs will have. If omitted, Hypothesis will choose one.
enable_vars: TODO(bjp): Make this `True` all the time and put variable
initialization in slicing_test. If `False`, the returned parameters are
all Tensors, never Variables or DeferredTensor.
depth: Python `int` giving maximum nesting depth of compound kernel.
Returns:
kernels: A strategy for drawing `SchurComplement` kernels with the specified
`batch_shape` (or an arbitrary one if omitted).
"""
if depth is None:
depth = draw(depths())
if batch_shape is None:
batch_shape = draw(tfp_hps.shapes())
if event_dim is None:
event_dim = draw(hps.integers(min_value=2, max_value=6))
if feature_dim is None:
feature_dim = draw(hps.integers(min_value=2, max_value=6))
if feature_ndims is None:
feature_ndims = draw(hps.integers(min_value=2, max_value=6))
base_kernel, kernel_variable_names = draw(
kernels(batch_shape=batch_shape,
event_dim=event_dim,
feature_dim=feature_dim,
feature_ndims=feature_ndims,
enable_vars=False,
depth=depth - 1))
# SchurComplement requires the inputs to have one example dimension.
fixed_inputs = draw(
kernel_input(batch_shape=batch_shape,
example_ndims=1,
feature_dim=feature_dim,
feature_ndims=feature_ndims))
# Positive shift to ensure the divisor matrix is PD.
diag_shift = np.float64(
draw(
hpnp.arrays(dtype=np.float64,
shape=tensorshape_util.as_list(batch_shape),
elements=hps.floats(1,
100,
allow_nan=False,
allow_infinity=False))))
hp.note('Forming SchurComplement kernel with fixed_inputs: {} '
'and diag_shift: {}'.format(fixed_inputs, diag_shift))
schur_complement_params = {
'fixed_inputs': fixed_inputs,
'diag_shift': diag_shift
}
for param_name in schur_complement_params:
if enable_vars and draw(hps.booleans()):
kernel_variable_names.append(param_name)
schur_complement_params[param_name] = tf.Variable(
schur_complement_params[param_name], name=param_name)
if draw(hps.booleans()):
schur_complement_params[
param_name] = tfp_hps.defer_and_count_usage(
schur_complement_params[param_name])
result_kernel = tfp.math.psd_kernels.SchurComplement(
base_kernel=base_kernel,
fixed_inputs=schur_complement_params['fixed_inputs'],
diag_shift=schur_complement_params['diag_shift'],
validate_args=True)
return result_kernel, kernel_variable_names
from newton2 import newton2
import numpy as np
import pytest
from hypothesis import given
import hypothesis.strategies as st
@given(st.floats(-10, 10), st.floats(-10, 10))
def test_unique_solution(x0, y0):
'''
tests if the algorithm correctly finds that the unique zero of the system
f(x,y) = x = 0
g(x,y) = y = 0
is (x,y)=(0,0). The strategy is to use different initial guesses for both x and y.
'''
def f1(x, y):
return x
def f2(x, y):
return y
def f(x, y):
return [f1(x, y), f2(x, y)]
def Jf(x, y):
return [[1, 0], [0, 1]]
p0 = [x0, y0]
xn, yn = newton2(f, Jf, p0).x
assert round(xn, 6) == 0
assert round(yn, 6) == 0
def kernel_input(draw,
batch_shape,
example_dim=None,
example_ndims=None,
feature_dim=None,
feature_ndims=None,
enable_vars=False,
name=None):
"""Strategy for drawing arbitrary Kernel input.
In order to avoid duplicates (or even numerically near-duplicates), we
generate inputs on a grid. We let hypothesis generate the number of grid
points and distance between grid points, within some reasonable pre-defined
ranges. The result will be a batch of example sets, within which each set of
examples has no duplicates (but no such duplication avoidance is applied
accross batches).
Args:
draw: Hypothesis function supplied by `@hps.composite`.
batch_shape: `TensorShape`. The batch shape of the resulting
kernel input.
example_dim: Optional Python int giving the size of each example dimension.
If omitted, Hypothesis will choose one.
example_ndims: Optional Python int giving the number of example dimensions
of the input. If omitted, Hypothesis will choose one.
feature_dim: Optional Python int giving the size of each feature dimension.
If omitted, Hypothesis will choose one.
feature_ndims: Optional Python int stating the number of feature dimensions
inputs will have. If omitted, Hypothesis will choose one.
enable_vars: If `False`, the returned parameters are all Tensors, never
Variables or DeferredTensor.
name: Name to give the variable.
Returns:
kernel_input: A strategy for drawing kernel_input with the prescribed shape
(or an arbitrary one if omitted).
"""
if example_ndims is None:
example_ndims = draw(hps.integers(min_value=1, max_value=2))
if example_dim is None:
example_dim = draw(hps.integers(min_value=2, max_value=4))
if feature_ndims is None:
feature_ndims = draw(hps.integers(min_value=1, max_value=2))
if feature_dim is None:
feature_dim = draw(hps.integers(min_value=2, max_value=4))
batch_shape = tensorshape_util.as_list(batch_shape)
example_shape = [example_dim] * example_ndims
feature_shape = [feature_dim] * feature_ndims
batch_size = int(np.prod(batch_shape))
example_size = example_dim**example_ndims
feature_size = feature_dim**feature_ndims
# We would like each batch of examples to be unique, to avoid computing kernel
# matrices that are semi-definite. hypothesis.extra.numpy.arrays doesn't have
# a sense of tolerance, so we need to do some extra work to get points
# sufficiently far from each other.
grid_size = draw(hps.integers(min_value=10, max_value=100))
grid_spacing = draw(hps.floats(min_value=1e-2, max_value=2))
hp.note('Grid size {} and spacing {}'.format(grid_size, grid_spacing))
def _grid_indices_to_values(grid_indices):
return (
grid_spacing *
(np.array(grid_indices, dtype=np.float64) - np.float64(grid_size)))
# We'll construct the result by stacking onto flattened batch, example and
# feature dims, then reshape to unflatten at the end.
result = np.zeros([0, example_size, feature_size])
for _ in range(batch_size):
seen = set()
index_array_strategy = hps.tuples(
*([hps.integers(0, grid_size + 1)] * feature_size)).filter(
lambda x, seen=seen: x not in seen
) # Default param to sate pylint.
examples = np.zeros([1, 0, feature_size])
for _ in range(example_size):
feature_grid_locations = draw(index_array_strategy)
seen.add(feature_grid_locations)
example = _grid_indices_to_values(feature_grid_locations)
example = example[np.newaxis, np.newaxis, ...]
examples = np.concatenate([examples, example], axis=1)
result = np.concatenate([result, examples], axis=0)
result = np.reshape(result, batch_shape + example_shape + feature_shape)
if enable_vars and draw(hps.booleans()):
result = tf.Variable(result, name=name)
if draw(hps.booleans()):
result = tfp_hps.defer_and_count_usage(result)
return result
from __future__ import absolute_import, division, print_function
import gc
import weakref
import pytest
import hypothesis.strategies as st
from hypothesis import given, settings
@pytest.mark.parametrize(
"s",
[
st.floats(),
st.tuples(st.integers()),
st.tuples(),
st.one_of(st.integers(), st.text()),
],
)
def test_is_cacheable(s):
assert s.is_cacheable
@pytest.mark.parametrize(
"s",
[
st.just([]),
st.tuples(st.integers(), st.just([])),
st.one_of(st.integers(), st.text(), st.just([])),
def sample_program_configs(self, draw):
in_shape = draw(
st.lists(st.integers(min_value=1, max_value=32),
min_size=4,
max_size=4))
is_test_val = draw(st.sampled_from([True, False]))
epsilon = draw(st.floats(min_value=0.00001, max_value=0.001))
momentum = draw(st.floats(min_value=0.1, max_value=0.9))
def generate_input(*args, **kwargs):
return np.random.random(in_shape).astype(np.float32)
def generate_scale(*args, **kwargs):
return np.random.random([in_shape[1]]).astype(np.float32) + 0.5
def generate_bias(*args, **kwargs):
return np.random.random([in_shape[1]]).astype(np.float32)
def generate_mean(*args, **kwargs):
return np.random.random([in_shape[1]]).astype(np.float32)
def generate_variance(*args, **kwargs):
return np.random.random([in_shape[1]]).astype(np.float32)
outputs = [
"output_data", "mean_data", "variance_data", "saved_mean",
"saved_variance"
]
if self.get_target() == "Metal":
outputs = ["output_data"]
batch_norm_ops = OpConfig(type="batch_norm",
inputs={
"X": ["input_data"],
"Scale": ["scale_data"],
"Bias": ["bias_data"],
"Mean": ["mean_data"],
"Variance": ["variance_data"]
},
outputs={
"Y": ["output_data"],
"MeanOut": ["mean_data"],
"VarianceOut": ["variance_data"],
"SavedMean": ["saved_mean"],
"SavedVariance": ["saved_variance"]
},
attrs={
"is_test": False,
"trainable_statistics": False,
"data_layout": "NCHW",
"use_global_stats": True,
"epsilon": epsilon,
"momentum": momentum
})
program_config = ProgramConfig(
ops=[batch_norm_ops],
weights={},
inputs={
"input_data": TensorConfig(data_gen=partial(generate_input)),
"scale_data": TensorConfig(data_gen=partial(generate_scale)),
"bias_data": TensorConfig(data_gen=partial(generate_bias)),
"mean_data": TensorConfig(data_gen=partial(generate_mean)),
"variance_data":
TensorConfig(data_gen=partial(generate_variance)),
},
outputs=outputs)
return program_config
class QCircuitMachine(RuleBasedStateMachine):
"""Build a Hypothesis rule based state machine for constructing, transpiling
and simulating a series of random QuantumCircuits.
Build circuits with up to QISKIT_RANDOM_QUBITS qubits, apply a random
selection of gates from qiskit.extensions.standard with randomly selected
qargs, cargs, and parameters. At random intervals, transpile the circuit for
a random backend with a random optimization level and simulate both the
initial and the transpiled circuits to verify that their counts are the
same.
"""
qubits = Bundle('qubits')
clbits = Bundle('clbits')
backend = Aer.get_backend('qasm_simulator')
max_qubits = int(backend.configuration().n_qubits / 2)
def __init__(self):
super().__init__()
self.qc = QuantumCircuit()
@precondition(lambda self: len(self.qc.qubits) < self.max_qubits)
@rule(target=qubits, n=st.integers(min_value=1, max_value=max_qubits))
def add_qreg(self, n):
"""Adds a new variable sized qreg to the circuit, up to max_qubits."""
n = min(n, self.max_qubits - len(self.qc.qubits))
qreg = QuantumRegister(n)
self.qc.add_register(qreg)
return multiple(*list(qreg))
@rule(target=clbits, n=st.integers(1, 5))
def add_creg(self, n):
"""Add a new variable sized creg to the circuit."""
creg = ClassicalRegister(n)
self.qc.add_register(creg)
return multiple(*list(creg))
# Gates of various shapes
@rule(gate=st.sampled_from(oneQ_gates), qarg=qubits)
def add_1q_gate(self, gate, qarg):
"""Append a random 1q gate on a random qubit."""
self.qc.append(gate(), [qarg], [])
@rule(gate=st.sampled_from(twoQ_gates),
qargs=st.lists(qubits, max_size=2, min_size=2, unique=True))
def add_2q_gate(self, gate, qargs):
"""Append a random 2q gate across two random qubits."""
self.qc.append(gate(), qargs)
@rule(gate=st.sampled_from(threeQ_gates),
qargs=st.lists(qubits, max_size=3, min_size=3, unique=True))
def add_3q_gate(self, gate, qargs):
"""Append a random 3q gate across three random qubits."""
self.qc.append(gate(), qargs)
@rule(gate=st.sampled_from(oneQ_oneP_gates),
qarg=qubits,
param=st.floats(allow_nan=False,
allow_infinity=False,
min_value=-10 * pi,
max_value=10 * pi))
def add_1q1p_gate(self, gate, qarg, param):
"""Append a random 1q gate with 1 random float parameter."""
self.qc.append(gate(param), [qarg])
@rule(gate=st.sampled_from(oneQ_twoP_gates),
qarg=qubits,
params=st.lists(st.floats(allow_nan=False,
allow_infinity=False,
min_value=-10 * pi,
max_value=10 * pi),
min_size=2,
max_size=2))
def add_1q2p_gate(self, gate, qarg, params):
"""Append a random 1q gate with 2 random float parameters."""
self.qc.append(gate(*params), [qarg])
@rule(gate=st.sampled_from(oneQ_threeP_gates),
qarg=qubits,
params=st.lists(st.floats(allow_nan=False,
allow_infinity=False,
min_value=-10 * pi,
max_value=10 * pi),
min_size=3,
max_size=3))
def add_1q3p_gate(self, gate, qarg, params):
"""Append a random 1q gate with 3 random float parameters."""
self.qc.append(gate(*params), [qarg])
@rule(gate=st.sampled_from(twoQ_oneP_gates),
qargs=st.lists(qubits, max_size=2, min_size=2, unique=True),
param=st.floats(allow_nan=False,
allow_infinity=False,
min_value=-10 * pi,
max_value=10 * pi))
def add_2q1p_gate(self, gate, qargs, param):
"""Append a random 2q gate with 1 random float parameter."""
self.qc.append(gate(param), qargs)
@rule(gate=st.sampled_from(twoQ_threeP_gates),
qargs=st.lists(qubits, max_size=2, min_size=2, unique=True),
params=st.lists(st.floats(allow_nan=False,
allow_infinity=False,
min_value=-10 * pi,
max_value=10 * pi),
min_size=3,
max_size=3))
def add_2q3p_gate(self, gate, qargs, params):
"""Append a random 2q gate with 3 random float parameters."""
self.qc.append(gate(*params), qargs)
@rule(gate=st.sampled_from(oneQ_oneC_gates), qarg=qubits, carg=clbits)
def add_1q1c_gate(self, gate, qarg, carg):
"""Append a random 1q, 1c gate."""
self.qc.append(gate(), [qarg], [carg])
@rule(gate=st.sampled_from(variadic_gates),
qargs=st.lists(qubits, min_size=1, unique=True))
def add_variQ_gate(self, gate, qargs):
"""Append a gate with a variable number of qargs."""
self.qc.append(gate(len(qargs)), qargs)
@precondition(lambda self: len(self.qc.data) > 0)
@rule(carg=clbits, data=st.data())
def add_c_if_last_gate(self, carg, data):
"""Modify the last gate to be conditional on a classical register."""
creg = carg.register
val = data.draw(st.integers(min_value=0, max_value=2**len(creg) - 1))
last_gate = self.qc.data[-1]
# Conditional instructions are not supported
assume(isinstance(last_gate[0], Gate))
last_gate[0].c_if(creg, val)
# Properties to check
@invariant()
def qasm(self):
"""After each circuit operation, it should be possible to build QASM."""
self.qc.qasm()
@precondition(
lambda self: any(isinstance(d[0], Measure) for d in self.qc.data))
@rule(backend=st.one_of(st.none(), st.sampled_from(mock_backends)),
opt_level=st.integers(min_value=0, max_value=3))
def equivalent_transpile(self, backend, opt_level):
"""Simulate, transpile and simulate the present circuit. Verify that the
counts are not significantly different before and after transpilation.
"""
print('Evaluating circuit at level {} on {}:\n{}'.format(
opt_level, backend, self.qc.qasm()))
assume(backend is None
or backend.configuration().n_qubits >= len(self.qc.qubits))
shots = 4096
aer_counts = execute(self.qc, backend=self.backend,
shots=shots).result().get_counts()
try:
xpiled_qc = transpile(self.qc,
backend=backend,
optimization_level=opt_level)
except Exception as e:
failed_qasm = 'Exception caught during transpilation of circuit: \n{}'.format(
self.qc.qasm())
raise RuntimeError(failed_qasm) from e
xpiled_aer_counts = execute(xpiled_qc,
backend=self.backend,
shots=shots).result().get_counts()
count_differences = dicts_almost_equal(aer_counts, xpiled_aer_counts,
0.05 * shots)
assert count_differences == '', 'Counts not equivalent: {}\nFailing QASM: \n{}'.format(
count_differences, self.qc.qasm())
import minitorch
import pytest
from hypothesis import given
from numba import cuda
from hypothesis.strategies import floats, integers, lists, data, permutations
from .strategies import tensors, shaped_tensors, assert_close
small_floats = floats(min_value=-100, max_value=100, allow_nan=False)
v = 4.524423
one_arg = [
("neg", lambda a: -a),
("addconstant", lambda a: a + v),
("lt", lambda a: a < v),
("subconstant", lambda a: a - v),
("mult", lambda a: 5 * a),
("div", lambda a: a / v),
("sig", lambda a: a.sigmoid()),
("log", lambda a: (a + 100000).log()),
("relu", lambda a: (a + 2).relu()),
("exp", lambda a: (a - 200).exp()),
]
reduce = [
("sum", lambda a: a.sum()),
("mean", lambda a: a.mean()),
("sum2", lambda a: a.sum(0)),
("mean2", lambda a: a.mean(0)),
]
two_arg = [
("add", lambda a, b: a + b),
("mul", lambda a, b: a * b),
class TestPairWiseLossOps(hu.HypothesisTestCase):
@given(X=hu.arrays(dims=[2, 1],
elements=st.floats(min_value=0.0, max_value=10.0)),
label=hu.arrays(dims=[2, 1],
elements=st.integers(min_value=0, max_value=1),
dtype=np.float32),
**hu.gcs_cpu_only)
def test_pair_wise_loss_predictions(self, X, label, gc, dc):
workspace.FeedBlob('X', X)
workspace.FeedBlob('label', label)
new_label = np.array([label[1], label[0]])
new_x = np.array([X[1], X[0]])
workspace.FeedBlob('new_x', new_x)
workspace.FeedBlob('new_label', new_label)
net = core.Net('net')
net.PairWiseLoss(['X', 'label'], ['output'])
net.PairWiseLoss(['new_x', 'new_label'], ['new_output'])
plan = core.Plan('predict_data')
plan.AddStep(core.execution_step('predict_data',
[net], num_iter=1))
workspace.RunPlan(plan)
output = workspace.FetchBlob('output')
new_output = workspace.FetchBlob('new_output')
sign = 1 if label[0] > label[1] else -1
if label[0] == label[1]:
self.assertEqual(np.asscalar(output), 0)
return
self.assertAlmostEqual(
np.asscalar(output),
np.asscalar(np.log(1 + np.exp(sign * (X[1] - X[0])))),
delta=1e-4
)
# check swapping row order doesn't alter overall loss
self.assertAlmostEqual(output, new_output)
@given(X=hu.arrays(dims=[2, 1],
elements=st.floats(min_value=0.0, max_value=10.0)),
label=hu.arrays(dims=[2, 1],
elements=st.integers(min_value=0, max_value=1),
dtype=np.float32),
dY=hu.arrays(dims=[1],
elements=st.floats(min_value=1, max_value=10)),
**hu.gcs_cpu_only)
def test_pair_wise_loss_gradient(self, X, label, dY, gc, dc):
workspace.FeedBlob('X', X)
workspace.FeedBlob('dY', dY)
workspace.FeedBlob('label', label)
net = core.Net('net')
net.PairWiseLossGradient(
['X', 'label', 'dY'],
['dX'],
)
plan = core.Plan('predict_data')
plan.AddStep(core.execution_step('predict_data',
[net], num_iter=1))
workspace.RunPlan(plan)
dx = workspace.FetchBlob('dX')
sign = 1 if label[0] > label[1] else -1
if label[0] == label[1]:
self.assertEqual(np.asscalar(dx[0]), 0)
return
self.assertAlmostEqual(
np.asscalar(dx[0]),
np.asscalar(-dY[0] * sign / (1 + np.exp(sign * (X[0] - X[1])))),
delta=1e-2 * abs(np.asscalar(dx[0])))
self.assertEqual(np.asscalar(dx[0]), np.asscalar(-dx[1]))
delta = 1e-3
up_x = np.array([[X[0] + delta], [X[1]]], dtype=np.float32)
down_x = np.array([[X[0] - delta], [X[1]]], dtype=np.float32)
workspace.FeedBlob('up_x', up_x)
workspace.FeedBlob('down_x', down_x)
new_net = core.Net('new_net')
new_net.PairWiseLoss(['up_x', 'label'], ['up_output'])
new_net.PairWiseLoss(['down_x', 'label'], ['down_output'])
plan = core.Plan('predict_data')
plan.AddStep(core.execution_step('predict_data', [new_net], num_iter=1))
workspace.RunPlan(plan)
down_output_pred = workspace.FetchBlob('down_output')
up_output_pred = workspace.FetchBlob('up_output')
np.testing.assert_allclose(
np.asscalar(dx[0]),
np.asscalar(
0.5 * dY[0] *
(up_output_pred[0] - down_output_pred[0]) / delta),
rtol=1e-2, atol=1e-2)
import pytest
from hypothesis import given, assume
from hypothesis.errors import UnsatisfiedAssumption
from hypothesis.strategies import integers, floats, one_of, just
from segpy.ibm_float import (ieee2ibm, ibm2ieee, MAX_IBM_FLOAT,
SMALLEST_POSITIVE_NORMAL_IBM_FLOAT,
LARGEST_NEGATIVE_NORMAL_IBM_FLOAT, MIN_IBM_FLOAT,
IBMFloat, EPSILON_IBM_FLOAT,
MAX_EXACT_INTEGER_IBM_FLOAT,
MIN_EXACT_INTEGER_IBM_FLOAT, EXPONENT_BIAS)
from segpy.util import almost_equal
ibm_compatible_negative_floats = floats(MIN_IBM_FLOAT,
LARGEST_NEGATIVE_NORMAL_IBM_FLOAT)
ibm_compatible_positive_floats = floats(SMALLEST_POSITIVE_NORMAL_IBM_FLOAT,
MAX_IBM_FLOAT)
ibm_compatible_non_negative_floats = one_of(
just(0.0), floats(SMALLEST_POSITIVE_NORMAL_IBM_FLOAT, MAX_IBM_FLOAT))
ibm_compatible_non_positive_floats = one_of(
just(0.0), floats(MIN_IBM_FLOAT, LARGEST_NEGATIVE_NORMAL_IBM_FLOAT))
def ibm_compatible_floats(min_f, max_f):
truncated_min_f = max(min_f, MIN_IBM_FLOAT)
truncated_max_f = min(max_f, MAX_IBM_FLOAT)
strategies = []
value = attr.ib() # type: object
deferred = attr.ib(attr.Factory(Deferred)) # type: Deferred
def resolve(self):
"""
Resolve the L{Deferred} that represents the value with the
value itself.
"""
self.deferred.callback(self.value)
jsonAtoms = (st.none()
| st.booleans()
| st.integers()
| st.floats(allow_nan=False)
| st.text(printable))
def jsonComposites(children):
"""
Creates a Hypothesis strategy that constructs composite
JSON-serializable objects (e.g., lists).
@param children: A strategy from which each composite object's
children will be drawn.
@return: The composite objects strategy.
"""
return (st.lists(children)
| st.dictionaries(st.text(printable), children)
dom0.firewall("test-vm", value, "192.168.1.1", "1")
# ~~~ dsthost ~~~
@hypothesis.given(s.one_of(s.integers(), s.text(), s.functions()))
def test__firewall__dsthost__fuzz_negative(value):
with patch("qsm.dom0.lib.run", return_value=None, autospec=True):
with patch("qsm.dom0.exists_or_throws",
return_value=True,
autospec=True):
with pytest.raises(AssertionError):
dom0.firewall("test-vm", "accept", value, "1")
# ~~~ dst ports ~~~
@hypothesis.given(s.one_of(s.floats(), s.text(), s.functions()))
def test__firewall__dstports__invalid_type_fuzz_negative(value):
"""Test random types cause assertion error"""
with patch("qsm.dom0.lib.run", return_value=None, autospec=True):
with patch("qsm.dom0.exists_or_throws",
return_value=True,
autospec=True):
with pytest.raises(AssertionError):
dom0.firewall("test-vm", "accept", "192.168.1.1", value)
@hypothesis.given(s.integers(min_value=1, max_value=65535))
def test__firewall__dstports__fuzz_positive(value):
"""Test values inside of the acceptable range of ports"""
with patch("qsm.dom0.lib.run", return_value=None, autospec=True):
with patch("qsm.dom0.exists_or_throws",
def bijectors(draw,
bijector_name=None,
batch_shape=None,
event_dim=None,
enable_vars=False):
"""Strategy for drawing Bijectors.
The emitted bijector may be a basic bijector or an `Invert` of a basic
bijector, but not a compound like `Chain`.
Args:
draw: Hypothesis strategy sampler supplied by `@hps.composite`.
bijector_name: Optional Python `str`. If given, the produced bijectors
will all have this type. If omitted, Hypothesis chooses one from
the whitelist `TF2_FRIENDLY_BIJECTORS`.
batch_shape: An optional `TensorShape`. The batch shape of the resulting
bijector. Hypothesis will pick one if omitted.
event_dim: Optional Python int giving the size of each of the underlying
distribution's parameters' event dimensions. This is shared across all
parameters, permitting square event matrices, compatible location and
scale Tensors, etc. If omitted, Hypothesis will choose one.
enable_vars: TODO(bjp): Make this `True` all the time and put variable
initialization in slicing_test. If `False`, the returned parameters are
all `tf.Tensor`s and not {`tf.Variable`, `tfp.util.DeferredTensor`
`tfp.util.TransformedVariable`}
Returns:
bijectors: A strategy for drawing bijectors with the specified `batch_shape`
(or an arbitrary one if omitted).
"""
if bijector_name is None:
bijector_name = draw(hps.sampled_from(TF2_FRIENDLY_BIJECTORS))
if batch_shape is None:
batch_shape = draw(tfp_hps.shapes())
if event_dim is None:
event_dim = draw(hps.integers(min_value=2, max_value=6))
if bijector_name == 'Invert':
underlying_name = draw(
hps.sampled_from(sorted(set(TF2_FRIENDLY_BIJECTORS) - {'Invert'})))
underlying = draw(
bijectors(bijector_name=underlying_name,
batch_shape=batch_shape,
event_dim=event_dim,
enable_vars=enable_vars))
return tfb.Invert(underlying, validate_args=True)
if bijector_name == 'TransformDiagonal':
underlying_name = draw(
hps.sampled_from(sorted(TRANSFORM_DIAGONAL_WHITELIST)))
underlying = draw(
bijectors(bijector_name=underlying_name,
batch_shape=(),
event_dim=event_dim,
enable_vars=enable_vars))
return tfb.TransformDiagonal(underlying, validate_args=True)
if bijector_name == 'Inline':
if enable_vars:
scale = tf.Variable(1., name='scale')
else:
scale = 2.
b = tfb.AffineScalar(scale=scale)
inline = tfb.Inline(
forward_fn=b.forward,
inverse_fn=b.inverse,
forward_log_det_jacobian_fn=lambda x: b.forward_log_det_jacobian( # pylint: disable=g-long-lambda
x,
event_ndims=b.forward_min_event_ndims),
forward_min_event_ndims=b.forward_min_event_ndims,
is_constant_jacobian=b.is_constant_jacobian,
)
inline.b = b
return inline
if bijector_name == 'DiscreteCosineTransform':
dct_type = draw(hps.integers(min_value=2, max_value=3))
return tfb.DiscreteCosineTransform(validate_args=True,
dct_type=dct_type)
if bijector_name == 'PowerTransform':
power = draw(hps.floats(min_value=0., max_value=10.))
return tfb.PowerTransform(validate_args=True, power=power)
if bijector_name == 'Permute':
event_ndims = draw(hps.integers(min_value=1, max_value=2))
axis = draw(hps.integers(min_value=-event_ndims, max_value=-1))
permutation = draw(hps.permutations(np.arange(event_dim)))
return tfb.Permute(permutation, axis=axis)
bijector_params = draw(
broadcasting_params(bijector_name,
batch_shape,
event_dim=event_dim,
enable_vars=enable_vars))
ctor = getattr(tfb, bijector_name)
return ctor(validate_args=True, **bijector_params)
standard_types = [
lists(max_size=0), tuples(), sets(max_size=0), frozensets(max_size=0),
fixed_dictionaries({}),
abc(booleans(), booleans(), booleans()),
abc(booleans(), booleans(), integers()),
fixed_dictionaries({'a': integers(), 'b': booleans()}),
dictionaries(booleans(), integers()),
dictionaries(text(), booleans()),
one_of(integers(), tuples(booleans())),
sampled_from(range(10)),
one_of(just('a'), just('b'), just('c')),
sampled_from(('a', 'b', 'c')),
integers(),
integers(min_value=3),
integers(min_value=(-2 ** 32), max_value=(2 ** 64)),
floats(), floats(min_value=-2.0, max_value=3.0),
floats(), floats(min_value=-2.0),
floats(), floats(max_value=-0.0),
floats(), floats(min_value=0.0),
floats(min_value=3.14, max_value=3.14),
text(), binary(),
booleans(),
tuples(booleans(), booleans()),
frozensets(integers()),
sets(frozensets(booleans())),
complex_numbers(),
fractions(),
decimals(),
lists(lists(booleans(), average_size=10), average_size=10),
lists(lists(booleans(), average_size=100)),
lists(floats(0.0, 0.0), average_size=1.0),
class TestActivations(serial.SerializedTestCase):
@serial.given(X=hu.tensor(), in_place=st.booleans(),
engine=st.sampled_from(["", "CUDNN"]), **mu.gcs)
def test_relu(self, X, in_place, engine, gc, dc):
if gc == mu.mkl_do:
in_place = False
op = core.CreateOperator(
"Relu",
["X"],
["X"] if in_place else ["Y"],
engine=engine,
)
def relu_ref(X):
return [np.maximum(X, 0.0)]
# go away from the origin point to avoid kink problems
X += 0.02 * np.sign(X)
X[X == 0.0] += 0.02
self.assertReferenceChecks(gc, op, [X], relu_ref)
self.assertDeviceChecks(dc, op, [X], [0])
self.assertGradientChecks(gc, op, [X], 0, [0])
@unittest.skipIf(not workspace.has_gpu_support and
not workspace.has_hip_support,
"Relu for float16 can only run on GPU now.")
@given(X=hu.tensor(dtype=np.float16), in_place=st.booleans(),
engine=st.sampled_from(["", "CUDNN"]), **hu.gcs)
def test_relu_fp16(self, X, in_place, engine, gc, dc):
# fp16 is only supported on CUDA/HIP
assume(core.IsGPUDeviceType(gc.device_type))
op = core.CreateOperator(
"Relu",
["X"],
["X"] if in_place else ["Y"],
engine=engine,
)
def relu_ref(X):
return [np.maximum(X, 0.0)]
def relu_grad_ref(g_out, outputs, fwd_inputs):
dY = g_out
[Y] = outputs
dX = dY
dX[Y == 0] = 0
return [dX]
# go away from the origin point to avoid kink problems
X += 0.02 * np.sign(X)
X[X == 0.0] += 0.02
self.assertReferenceChecks(
gc,
op,
[X],
relu_ref,
output_to_grad="X" if in_place else "Y",
grad_reference=relu_grad_ref)
@serial.given(X=hu.tensor(elements=st.floats(-3.0, 3.0)),
n=st.floats(min_value=0.5, max_value=2.0),
in_place=st.booleans(), **hu.gcs)
def test_relu_n(self, X, n, in_place, gc, dc):
op = core.CreateOperator(
"ReluN",
["X"],
["X"] if in_place else ["Y"],
n=n,
)
def relu_n_ref(X):
return [np.minimum(np.maximum(X, 0.0), n)]
# go away from 0 and n to avoid kink problems
X += 0.04 * np.sign(X)
X[X == 0.0] += 0.04
X -= n
X += 0.02 * np.sign(X)
X[X == 0.0] -= 0.02
X += n
self.assertReferenceChecks(gc, op, [X], relu_n_ref)
self.assertDeviceChecks(dc, op, [X], [0])
self.assertGradientChecks(gc, op, [X], 0, [0], stepsize=0.005)
@serial.given(X=hu.tensor(),
alpha=st.floats(min_value=0.1, max_value=2.0),
in_place=st.booleans(), engine=st.sampled_from(["", "CUDNN"]),
**hu.gcs)
def test_elu(self, X, alpha, in_place, engine, gc, dc):
op = core.CreateOperator(
"Elu",
["X"],
["X"] if in_place else ["Y"],
alpha=alpha,
engine=engine,
)
def elu_ref(X):
Y = X
Y[X < 0] = alpha * (np.exp(X[X < 0]) - 1.0)
return [Y]
# go away from the origin point to avoid kink problems
X += 0.04 * np.sign(X)
X[X == 0.0] += 0.04
self.assertReferenceChecks(gc, op, [X], elu_ref)
self.assertDeviceChecks(dc, op, [X], [0])
self.assertGradientChecks(gc, op, [X], 0, [0], stepsize=1e-2)
@given(X=hu.tensor(min_dim=4, max_dim=4),
alpha=st.floats(min_value=0.1, max_value=2.0),
inplace=st.booleans(),
shared=st.booleans(),
order=st.sampled_from(["NCHW", "NHWC"]),
seed=st.sampled_from([20, 100]),
**hu.gcs)
def test_prelu(self, X, alpha, inplace, shared, order, seed, gc, dc):
np.random.seed(seed)
W = np.random.randn(
X.shape[1] if order == "NCHW" else X.shape[3]).astype(np.float32)
if shared:
W = np.random.randn(1).astype(np.float32)
# go away from the origin point to avoid kink problems
X += 0.04 * np.sign(X)
X[X == 0.0] += 0.04
def prelu_ref(X, W):
Y = X.copy()
W = W.reshape(1, -1, 1, 1) if order == "NCHW" \
else W.reshape(1, 1, 1, -1)
assert len(X.shape) == 4
neg_indices = X <= 0
assert len(neg_indices.shape) == 4
assert X.shape == neg_indices.shape
Y[neg_indices] = (Y * W)[neg_indices]
return (Y,)
op = core.CreateOperator(
"PRelu", ["X", "W"], ["Y" if not inplace else "X"],
alpha=alpha, order=order)
self.assertReferenceChecks(gc, op, [X, W], prelu_ref)
# Check over multiple devices
self.assertDeviceChecks(dc, op, [X, W], [0])
if not inplace:
# Gradient check wrt X
self.assertGradientChecks(gc, op, [X, W], 0, [0], stepsize=1e-2)
# Gradient check wrt W
self.assertGradientChecks(gc, op, [X, W], 1, [0], stepsize=1e-2)
@serial.given(X=hu.tensor(),
alpha=st.floats(min_value=0.1, max_value=2.0),
inplace=st.booleans(),
**hu.gcs)
def test_leaky_relu(self, X, alpha, inplace, gc, dc):
# go away from the origin point to avoid kink problems
X += 0.04 * np.sign(X)
X[X == 0.0] += 0.04
def leaky_relu_ref(X):
Y = X.copy()
neg_indices = X <= 0
Y[neg_indices] = Y[neg_indices] * alpha
return (Y,)
op = core.CreateOperator(
"LeakyRelu",
["X"], ["Y" if not inplace else "X"],
alpha=alpha)
self.assertReferenceChecks(gc, op, [X], leaky_relu_ref)
# Check over multiple devices
self.assertDeviceChecks(dc, op, [X], [0])
@given(X=hu.tensor(),
inplace=st.booleans(),
**hu.gcs)
def test_leaky_relu_default(self, X, inplace, gc, dc):
# go away from the origin point to avoid kink problems
X += 0.04 * np.sign(X)
X[X == 0.0] += 0.04
def leaky_relu_ref(X):
Y = X.copy()
neg_indices = X <= 0
Y[neg_indices] = Y[neg_indices] * 0.01
return (Y,)
op = core.CreateOperator(
"LeakyRelu",
["X"], ["Y" if not inplace else "X"])
self.assertReferenceChecks(gc, op, [X], leaky_relu_ref)
# Check over multiple devices
self.assertDeviceChecks(dc, op, [X], [0])
class TestMatch(unittest.TestCase):
@given(turns=integers(min_value=1, max_value=200), game=games())
@example(turns=5, game=axl.DefaultGame)
def test_init(self, turns, game):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), turns, game=game)
self.assertEqual(match.result, [])
self.assertEqual(match.players, [p1, p2])
self.assertEqual(match.turns, turns)
self.assertEqual(match.prob_end, 0)
self.assertEqual(match.noise, 0)
self.assertEqual(match.game.RPST(), game.RPST())
self.assertEqual(match.players[0].match_attributes["length"], turns)
self.assertEqual(match._cache, {})
@given(prob_end=floats(min_value=0, max_value=1), game=games())
def test_init_with_prob_end(self, prob_end, game):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), prob_end=prob_end, game=game)
self.assertEqual(match.result, [])
self.assertEqual(match.players, [p1, p2])
self.assertEqual(match.turns, float("inf"))
self.assertEqual(match.prob_end, prob_end)
self.assertEqual(match.noise, 0)
self.assertEqual(match.game.RPST(), game.RPST())
self.assertEqual(match.players[0].match_attributes["length"],
float("inf"))
self.assertEqual(match._cache, {})
@given(
prob_end=floats(min_value=0, max_value=1),
turns=integers(min_value=1, max_value=200),
game=games(),
)
def test_init_with_prob_end_and_turns(self, turns, prob_end, game):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), turns=turns, prob_end=prob_end, game=game)
self.assertEqual(match.result, [])
self.assertEqual(match.players, [p1, p2])
self.assertEqual(match.turns, turns)
self.assertEqual(match.prob_end, prob_end)
self.assertEqual(match.noise, 0)
self.assertEqual(match.game.RPST(), game.RPST())
self.assertEqual(match.players[0].match_attributes["length"],
float("inf"))
self.assertEqual(match._cache, {})
def test_default_init(self):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2))
self.assertEqual(match.result, [])
self.assertEqual(match.players, [p1, p2])
self.assertEqual(match.turns, axl.DEFAULT_TURNS)
self.assertEqual(match.prob_end, 0)
self.assertEqual(match.noise, 0)
self.assertEqual(match.game.RPST(), (3, 1, 0, 5))
self.assertEqual(match.players[0].match_attributes["length"],
axl.DEFAULT_TURNS)
self.assertEqual(match._cache, {})
def test_example_prob_end(self):
"""
Test that matches have diff length and also that cache has recorded the
outcomes
"""
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), prob_end=0.5)
expected_lengths = [3, 1, 5]
for seed, expected_length in zip(range(3), expected_lengths):
axl.seed(seed)
self.assertEqual(match.players[0].match_attributes["length"],
float("inf"))
self.assertEqual(len(match.play()), expected_length)
self.assertEqual(match.noise, 0)
self.assertEqual(match.game.RPST(), (3, 1, 0, 5))
self.assertEqual(len(match._cache), 1)
self.assertEqual(match._cache[(p1, p2)], [(C, C)] * 5)
@given(turns=integers(min_value=1, max_value=200), game=games())
@example(turns=5, game=axl.DefaultGame)
def test_non_default_attributes(self, turns, game):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match_attributes = {"length": 500, "game": game, "noise": 0.5}
match = axl.Match((p1, p2),
turns,
game=game,
match_attributes=match_attributes)
self.assertEqual(match.players[0].match_attributes["length"], 500)
self.assertEqual(match.players[0].match_attributes["noise"], 0.5)
@given(turns=integers(min_value=1, max_value=200))
@example(turns=5)
def test_len(self, turns):
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), turns)
self.assertEqual(len(match), turns)
def test_len_error(self):
"""
Length is not defined if it is infinite.
"""
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), prob_end=0.5)
with self.assertRaises(TypeError):
len(match)
@given(p=floats(min_value=0, max_value=1))
def test_stochastic(self, p):
assume(0 < p < 1)
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), 5)
self.assertFalse(match._stochastic)
match = axl.Match((p1, p2), 5, noise=p)
self.assertTrue(match._stochastic)
p1 = axl.Random()
match = axl.Match((p1, p2), 5)
self.assertTrue(match._stochastic)
@given(p=floats(min_value=0, max_value=1))
def test_cache_update_required(self, p):
assume(0 < p < 1)
p1, p2 = axl.Cooperator(), axl.Cooperator()
match = axl.Match((p1, p2), 5, noise=p)
self.assertFalse(match._cache_update_required)
cache = DeterministicCache()
cache.mutable = False
match = axl.Match((p1, p2), 5, deterministic_cache=cache)
self.assertFalse(match._cache_update_required)
match = axl.Match((p1, p2), 5)
self.assertTrue(match._cache_update_required)
p1 = axl.Random()
match = axl.Match((p1, p2), 5)
self.assertFalse(match._cache_update_required)
def test_play(self):
cache = DeterministicCache()
players = (axl.Cooperator(), axl.Defector())
match = axl.Match(players, 3, deterministic_cache=cache)
expected_result = [(C, D), (C, D), (C, D)]
self.assertEqual(match.play(), expected_result)
self.assertEqual(cache[(axl.Cooperator(), axl.Defector())],
expected_result)
# a deliberately incorrect result so we can tell it came from the cache
expected_result = [(C, C), (D, D), (D, C), (C, C), (C, D)]
cache[(axl.Cooperator(), axl.Defector())] = expected_result
match = axl.Match(players, 3, deterministic_cache=cache)
self.assertEqual(match.play(), expected_result[:3])
def test_cache_grows(self):
"""
We want to make sure that if we try to use the cache for more turns than
what is stored, then it will instead regenerate the result and overwrite
the cache.
"""
cache = DeterministicCache()
players = (axl.Cooperator(), axl.Defector())
match = axl.Match(players, 3, deterministic_cache=cache)
expected_result_5_turn = [(C, D), (C, D), (C, D), (C, D), (C, D)]
expected_result_3_turn = [(C, D), (C, D), (C, D)]
self.assertEqual(match.play(), expected_result_3_turn)
match.turns = 5
self.assertEqual(match.play(), expected_result_5_turn)
# The cache should now hold the 5-turn result..
self.assertEqual(cache[(axl.Cooperator(), axl.Defector())],
expected_result_5_turn)
def test_cache_doesnt_shrink(self):
"""
We want to make sure that when we access the cache looking for fewer
turns than what is stored, then it will not overwrite the cache with the
shorter result.
"""
cache = DeterministicCache()
players = (axl.Cooperator(), axl.Defector())
match = axl.Match(players, 5, deterministic_cache=cache)
expected_result_5_turn = [(C, D), (C, D), (C, D), (C, D), (C, D)]
expected_result_3_turn = [(C, D), (C, D), (C, D)]
self.assertEqual(match.play(), expected_result_5_turn)
match.turns = 3
self.assertEqual(match.play(), expected_result_3_turn)
# The cache should still hold the 5.
self.assertEqual(cache[(axl.Cooperator(), axl.Defector())],
expected_result_5_turn)
def test_scores(self):
player1 = axl.TitForTat()
player2 = axl.Defector()
match = axl.Match((player1, player2), 3)
self.assertEqual(match.scores(), [])
match.play()
self.assertEqual(match.scores(), [(0, 5), (1, 1), (1, 1)])
def test_final_score(self):
player1 = axl.TitForTat()
player2 = axl.Defector()
match = axl.Match((player1, player2), 3)
self.assertEqual(match.final_score(), None)
match.play()
self.assertEqual(match.final_score(), (2, 7))
match = axl.Match((player2, player1), 3)
self.assertEqual(match.final_score(), None)
match.play()
self.assertEqual(match.final_score(), (7, 2))
def test_final_score_per_turn(self):
turns = 3
player1 = axl.TitForTat()
player2 = axl.Defector()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.final_score_per_turn(), None)
match.play()
self.assertEqual(match.final_score_per_turn(), (2 / turns, 7 / turns))
match = axl.Match((player2, player1), turns)
self.assertEqual(match.final_score_per_turn(), None)
match.play()
self.assertEqual(match.final_score_per_turn(), (7 / turns, 2 / turns))
def test_winner(self):
player1 = axl.TitForTat()
player2 = axl.Defector()
match = axl.Match((player1, player2), 3)
self.assertEqual(match.winner(), None)
match.play()
self.assertEqual(match.winner(), player2)
match = axl.Match((player2, player1), 3)
self.assertEqual(match.winner(), None)
match.play()
self.assertEqual(match.winner(), player2)
player1 = axl.Defector()
match = axl.Match((player1, player2), 3)
self.assertEqual(match.winner(), None)
match.play()
self.assertEqual(match.winner(), False)
def test_cooperation(self):
turns = 3
player1 = axl.Cooperator()
player2 = axl.Alternator()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.cooperation(), None)
match.play()
self.assertEqual(match.cooperation(), (3, 2))
player1 = axl.Alternator()
player2 = axl.Defector()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.cooperation(), None)
match.play()
self.assertEqual(match.cooperation(), (2, 0))
def test_normalised_cooperation(self):
turns = 3
player1 = axl.Cooperator()
player2 = axl.Alternator()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.normalised_cooperation(), None)
match.play()
self.assertEqual(match.normalised_cooperation(),
(3 / turns, 2 / turns))
player1 = axl.Alternator()
player2 = axl.Defector()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.normalised_cooperation(), None)
match.play()
self.assertEqual(match.normalised_cooperation(),
(2 / turns, 0 / turns))
def test_state_distribution(self):
turns = 3
player1 = axl.Cooperator()
player2 = axl.Alternator()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.state_distribution(), None)
match.play()
expected = Counter({(C, C): 2, (C, D): 1})
self.assertEqual(match.state_distribution(), expected)
player1 = axl.Alternator()
player2 = axl.Defector()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.state_distribution(), None)
match.play()
expected = Counter({(C, D): 2, (D, D): 1})
self.assertEqual(match.state_distribution(), expected)
def test_normalised_state_distribution(self):
turns = 3
player1 = axl.Cooperator()
player2 = axl.Alternator()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.normalised_state_distribution(), None)
match.play()
expected = Counter({(C, C): 2 / turns, (C, D): 1 / turns})
self.assertEqual(match.normalised_state_distribution(), expected)
player1 = axl.Alternator()
player2 = axl.Defector()
match = axl.Match((player1, player2), turns)
self.assertEqual(match.normalised_state_distribution(), None)
match.play()
expected = Counter({(C, D): 2 / turns, (D, D): 1 / turns})
self.assertEqual(match.normalised_state_distribution(), expected)
def test_sparklines(self):
players = (axl.Cooperator(), axl.Alternator())
match = axl.Match(players, 4)
match.play()
expected_sparklines = "████\n█ █ "
self.assertEqual(match.sparklines(), expected_sparklines)
expected_sparklines = "XXXX\nXYXY"
self.assertEqual(match.sparklines("X", "Y"), expected_sparklines)
def sample_program_configs(self, draw):
in_shape = draw(
st.lists(st.integers(min_value=1, max_value=20),
min_size=2,
max_size=5))
fill_constant_shape = draw(
st.lists(st.integers(min_value=1, max_value=20),
min_size=2,
max_size=5))
axis = draw(
st.integers(min_value=-1,
max_value=max(len(in_shape),
len(fill_constant_shape))))
out_shape = []
assume(
check_input_shape_available(in_shape_x=in_shape,
in_shape_y=fill_constant_shape,
axis=axis,
out_shape=out_shape) == True)
assume(out_shape == in_shape)
threshold = draw(st.floats(min_value=0, max_value=1))
scale = draw(st.floats(min_value=0.5, max_value=5))
offset = draw(st.floats(min_value=0, max_value=1))
hard_swish_op0 = OpConfig(type="hard_swish",
inputs={"X": ["input_data"]},
outputs={"Out": ["hard_swish_output_data"]},
attrs={
"threshold": threshold,
"scale": scale,
"offset": offset
})
fill_constant_op = OpConfig(
type="fill_constant",
inputs={},
outputs={"Out": ["fill_constant_output_data"]},
attrs={
"dtype": 5,
"shape": fill_constant_shape,
"value": 1.,
"force_cpu": False,
"place_type": -1
})
elementwise_mul_op = OpConfig(
type="elementwise_mul",
inputs={
"X": ["hard_swish_output_data"],
"Y": ["fill_constant_output_data"]
},
outputs={"Out": ["elementwise_mul_output_data"]},
attrs={"axis": axis})
hard_swish_op1 = OpConfig(
type="hard_swish",
inputs={"X": ["elementwise_mul_output_data"]},
outputs={"Out": ["output_data"]},
attrs={
"threshold": threshold,
"scale": scale,
"offset": offset
})
ops = [
hard_swish_op0, fill_constant_op, elementwise_mul_op,
hard_swish_op1
]
program_config = ProgramConfig(
ops=ops,
weights={},
inputs={"input_data": TensorConfig(shape=in_shape)},
outputs=["output_data"])
return program_config
def ndarrays_of_shape(shape, lo=-1000.0, hi=1000.0):
return arrays('float32', shape=shape, elements=floats(min_value=lo, max_value=hi))
""" test primitive space contains, discretizing and sampling """
import hypothesis as hp
import hypothesis.strategies as st
import pytest
from energypy.common.spaces import ContinuousSpace, DiscreteSpace
@hp.given(st.floats(min_value=-100, max_value=0),
st.floats(min_value=1, max_value=100), st.integers(1, 10))
def test_continuous(low, high, num_discrete):
space = ContinuousSpace('state', low, high, None)
assert space.low == low
assert space.high == high
assert space.contains(low)
assert space.contains((low + high / 2))
assert space.contains(high)
with pytest.raises(ValueError):
assert space.contains(low - 2)
assert space.contains(high + 2)
discrete = space.discretize(num_discrete)
assert len(discrete) == num_discrete
@hp.given(
st.integers(min_value=6, max_value=100), )
Test properties of the coordinate transforms in the field vector module to
test if everything has been correctly implemented.
"""
import numpy as np
import json
from hypothesis import given, settings
from hypothesis.strategies import floats
from hypothesis.strategies import tuples
from qcodes.math.field_vector import FieldVector
from qcodes.utils.helpers import NumpyJSONEncoder
random_coordinates = {
"cartesian":
tuples(
floats(min_value=0, max_value=1), # x
floats(min_value=0, max_value=1), # y
floats(min_value=0, max_value=1) # z
),
"spherical":
tuples(
floats(min_value=0, max_value=1), # r
floats(min_value=0, max_value=180), # theta
floats(min_value=0, max_value=180) # phi
),
"cylindrical":
tuples(
floats(min_value=0, max_value=1), # rho
floats(min_value=0, max_value=180), # phi
floats(min_value=0, max_value=1) # z
)
data,
)
from cattr import Converter
from . import (
primitive_strategies,
seqs_of_primitives,
lists_of_primitives,
dicts_of_primitives,
enums_of_primitives,
)
from ._compat import change_type_param
ints_and_type = tuples(integers(), just(int))
floats_and_type = tuples(floats(allow_nan=False), just(float))
strs_and_type = tuples(text(), just(unicode))
bytes_and_type = tuples(binary(), just(bytes))
primitives_and_type = one_of(ints_and_type, floats_and_type, strs_and_type,
bytes_and_type)
mut_set_types = sampled_from([Set, MutableSet])
set_types = one_of(mut_set_types, just(FrozenSet))
def create_generic_type(generic_types, param_type):
"""Create a strategy for generating parameterized generic types."""
return one_of(
generic_types,
generic_types.map(lambda t: t[Any]),
def assert_allclose_na(a, b):
"""assert_allclose with a broader NA/nan/None definition."""
if _is_na(a) and _is_na(b):
pass
else:
npt.assert_allclose(a, b)
@pytest.mark.parametrize("op, pandas_op", [(sum_op, pd.Series.sum),
(prod_op, pd.Series.prod)])
@settings(deadline=timedelta(milliseconds=1000))
@given(
data=st.lists(
st.one_of(st.floats(max_value=10.0, min_value=-10), st.none())),
skipna=st.booleans(),
)
def test_reduce_op(data, skipna, op, pandas_op):
arrow = pa.array(data, type=pa.float64(), from_pandas=True)
pandas = pd.Series(data, dtype=float)
assert_allclose_na(op(arrow, skipna), pandas_op(pandas, skipna=skipna))
# Split in the middle and check whether this still works
if len(data) > 2:
arrow = pa.chunked_array([
pa.array(data[:len(data) // 2],
type=pa.float64(),
from_pandas=True),
pa.array(data[len(data) // 2:],
import numpy as np
import pytest
from hypothesis import given
from hypothesis.extra.numpy import arrays
from hypothesis.strategies import floats
from pytest import approx
from coxeter.shape_classes.circle import Circle
@given(floats(0.1, 1000))
def test_perimeter(r):
circle = Circle(1)
circle.radius = r
assert circle.perimeter == 2 * np.pi * r
assert circle.circumference == 2 * np.pi * r
@given(floats(0.1, 1000))
def test_area(r):
circle = Circle(1)
circle.radius = r
assert circle.area == np.pi * r**2
@given(floats(0.1, 1000))
def test_set_area(area):
"""Test setting the area."""
circle = Circle(1)
circle.area = area
assert circle.area == approx(area)
def test_caches_floats_sensitively():
assert st.floats(min_value=0.0) is st.floats(min_value=0.0)
assert st.floats(min_value=0.0) is not st.floats(min_value=0)
assert st.floats(min_value=0.0) is not st.floats(min_value=-0.0)
def _check_style(style: str) -> None:
if style not in ("pytest", "unittest"):
raise InvalidArgument(f"Valid styles are 'pytest' or 'unittest', got {style!r}")
# Simple strategies to guess for common argument names - we wouldn't do this in
# builds() where strict correctness is required, but we only use these guesses
# when the alternative is nothing() to force user edits anyway.
#
# This table was constructed manually after skimming through the documentation
# for the builtins and a few stdlib modules. Future enhancements could be based
# on analysis of type-annotated code to detect arguments which almost always
# take values of a particular type.
_GUESS_STRATEGIES_BY_NAME = (
(st.text(), ["name", "filename", "fname"]),
(st.floats(), ["real", "imag"]),
(st.functions(), ["function", "func", "f"]),
(st.iterables(st.integers()) | st.iterables(st.text()), ["iterable"]),
)
def _strategy_for(param: inspect.Parameter) -> Union[st.SearchStrategy, InferType]:
# If our default value is an Enum or a boolean, we assume that any value
# of that type is acceptable. Otherwise, we only generate the default.
if isinstance(param.default, bool):
return st.booleans()
if isinstance(param.default, enum.Enum):
return st.sampled_from(type(param.default))
if param.default is not inspect.Parameter.empty:
# Using `st.from_type(type(param.default))` would introduce spurious
# failures in cases like the `flags` argument to regex functions.
