def get_updated_or_empty_result(use_updated_result, seed):
if use_updated_result:
return test_update(seed)
else:
result = results.Result()
result.begin()
return result
def test_update(seed, result=None):
"""
Test the update method of the Result class. Also return the result, which
can be used as an input to other test functions. Can accept a pre-existing
Result object, EG if it has been initialised with a specific output file
"""
np.random.seed(seed)
output_dim = np.random.randint(2, 5)
n = get_random_network(input_dim=2, output_dim=output_dim)
d = data.Sinusoidal(input_dim=2, output_dim=output_dim, n_train=100)
w = n.get_parameter_vector()
if result is None:
result = results.Result()
result.begin()
# Call the result.update method a few times
for i in range(5):
n.set_parameter_vector(w + i)
result.update(model=n, dataset=d, iteration=i)
# Check that value lists for the default columns have non-zero length
assert len(result.get_values(optimisers.results.columns.TrainError)) > 0
assert len(result.get_values(optimisers.results.columns.TestError)) > 0
assert len(result.get_values(optimisers.results.columns.Iteration)) > 0
assert len(result.get_values(optimisers.results.columns.Time)) > 0
return result
def test_get_iteration_number():
""" Unit test for the public get_iteration_number method of the Result class
"""
# Create result object, and tell it to begin
result = results.Result(add_default_columns=False)
result.begin()
# Check the initial iteration number is 0
assert result.get_iteration_number() == 0
# Check that the iteration number is updated when calling update
for i in range(10):
result.update(iteration=i)
assert result.get_iteration_number() == i
def test_output_file():
"""
Test the methods of a Result object, writing the outputs to a text file
"""
seed = 8151
np.random.seed(seed)
filename = "Result tests output.txt"
path = os.path.join(output_dir, filename)
with open(path, "w") as f:
result = results.Result("Test result", True, f)
test_update(seed=seed, result=result)
result.display_headers()
result._display_last()
result.display_summary(10)
def test_line_search_column(seed):
"""
Test that changes in the step size attribute of a LineSearch object are
reflected in the value lists of a Result object when the Result object is
updated.
"""
np.random.seed(seed)
num_steps = np.random.randint(10, 20)
step_sizes = np.random.uniform(0, 10, num_steps)
filename = "Test StepSize results column.txt"
path = os.path.join(output_dir, filename)
with open(path, "w") as f:
result = results.Result(add_default_columns=False, file=f)
ls = LineSearch()
col = results.columns.StepSize(ls)
result.add_column(col)
result.begin()
for s in step_sizes:
ls.s = s
result.update()
assert np.all(result.get_values(type(col)) == step_sizes)