def test_happy_path(self):
decoder = DFS(MockSampler())
results = list(decoder("x0"))
assert results == [
Result("0", 3.0, True, 1),
Result("10", 5.0, True, 1),
Result("110", 7.0, True, 1),
Result("20", 4.0, True, 1),
Result("210", 6.0, True, 1)
]
def __call__(self, input: Dict[str, Any], n_required_output=None):
for i, value in enumerate(self.values):
time.sleep(2)
yield Result(value, 1.0 / (i + 1), True, 1)
def __call__(self, input, n_required_output=None):
n_required_output = n_required_output or 1
for _ in range(n_required_output):
yield Result(input["value"], 0, True, 1)
def test_abort(self):
decoder = MockBeamSearch(3, 2)
results = list(decoder("".join([" "] * 100)))
assert [Result("0", -1.0, True, 1)] == results
def test_not_finished_output(self):
decoder = MockBeamSearch(3, 2, False)
results = list(decoder("x0"))
assert [Result("0", -1.0, False, 1), Result("x0", 0.0, True, 1),
Result("00", -1.0, True, 1), Result("10", -2.0, True, 1)
] == results
def test_happy_path(self):
decoder = MockBeamSearch(3, 100)
results = list(decoder("x0"))
assert [Result("0", -1.0, True, 1), Result("x0", 0.0, True, 1),
Result("10", -2.0, True, 1)] == results
def __call__(self, input: Environment, n_required_output=None):
y = self.model(input)["y"]
for _ in range(n_required_output):
out = int(torch.normal(mean=y, std=5).item())
yield Result(out, torch.abs(out - y).item(), True, 1)
def __call__(self, input: Environment, n_required_output=None):
for i, ast in enumerate(self.asts):
yield Result(ast, 1.0 / (i + 1), True, 1)