-
Notifications
You must be signed in to change notification settings - Fork 21
Open
Description
Hello!
Thanks for developping benchopt!
Would it be possible to add a unit test in the template?
Why:
- ease debugging base functions (typically this would ease interactive debugging),
- give a good idea of the flow directly in the template.
For example one could add a file 'test_pipeline.py' with something like.
from datasets.simulated import Dataset
from objective import Objective
from solvers.python-gd import Solver
# This test requires having installed the dependencies before
def test_pipeline():
# We add manually attributes of dataset, objective, solver for the test
dataset = Dataset()
dataset.n_samples = 1000
dataset.n_features = 500
dataset.random_state = 27
dataset_params = dataset.get_data()
objective = Objective()
objective.whiten_y = True
objective.set_data(**dataset_params)
objective_params = objective.get_objective()
solver = Solver()
solver.scale_step = 1.
solver.set_objective(**objective_params)
solver.run(n_iter=10)
results = solver.get_result()
metrics = objective.compute(*results)
assert 'value' in metrics.keys()
Metadata
Metadata
Assignees
Labels
No labels