def test_e2e():
hist = History([[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5],
[torch.tensor(2), 2, 5]],
[[torch.tensor(1), 0, 1], [torch.tensor(1), 1, 5]])
desired_dict_lhs = defaultdict(list)
desired_dict_lhs[0] = [[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5]]
desired_dict_lhs[2] = [[torch.tensor(2), 2, 5]]
desired_dict_rhs = defaultdict(list)
desired_dict_rhs[0] = [[torch.tensor(1), 0, 1]]
desired_dict_rhs[1] = [[torch.tensor(1), 1, 5]]
currpool = CurrentElems([[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5]],
[[torch.tensor(1), 0, 1]])
e_weights_type = toy_e_weights_type()
potentials = torch.tensor([0.0, 0.0, 0.0, 0.0, 0.0])
match_edges, e_weights_full = compute_matching(currpool, potentials,
e_weights_type)
result_pool, total_loss = step_simulation(currpool, match_edges,
e_weights_full, desired_dict_lhs,
desired_dict_rhs, 1)
assert approx(total_loss, 0.1)
assert result_pool.lhs == [[torch.tensor(2), 0, 5]]
assert result_pool.rhs == [[torch.tensor(1), 1, 5]]
def test_step_simulation():
hist = History([[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5],
[torch.tensor(2), 2, 5]],
[[torch.tensor(1), 0, 1], [torch.tensor(1), 1, 5]])
desired_dict_lhs = defaultdict(list)
desired_dict_lhs[0] = [[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5]]
desired_dict_lhs[2] = [[torch.tensor(2), 2, 5]]
desired_dict_rhs = defaultdict(list)
desired_dict_rhs[0] = [[torch.tensor(1), 0, 1]]
desired_dict_rhs[1] = [[torch.tensor(1), 1, 5]]
currpool = CurrentElems([[torch.tensor(1), 0, 2], [torch.tensor(2), 0, 5]],
[[torch.tensor(1), 0, 1]])
match_edges = torch.tensor([[0.0], [1.0]])
e_weights_type = toy_e_weights_type()
e_weights = weight_matrix(currpool.lhs, currpool.rhs, e_weights_type)
result_pool, total_loss = step_simulation(currpool, match_edges, e_weights,
desired_dict_lhs,
desired_dict_rhs, 1)
assert approx(total_loss, 0.1)
assert result_pool.lhs == [[torch.tensor(1), 0, 2]]
assert result_pool.rhs == [[torch.tensor(1), 1, 5]]
def unambiguous_matching():
currpool = CurrentElems([[torch.tensor(2), 0, 5], [torch.tensor(1), 0, 5], [torch.tensor(2), 0, 5]],
[[torch.tensor(0), 0, 5]])
e_weights = toy_e_weights_type()
correct_matching = torch.tensor([[0.0],
[1.0],
[0.0]])
return currpool, e_weights, correct_matching
def test_weight_matrix():
currpool = CurrentElems([[torch.tensor(2), 0, 5], [torch.tensor(1), 0, 5], [torch.tensor(2), 0, 5]],
[[torch.tensor(0), 0, 5]])
e_weights = toy_e_weights_type()
result_weights = weight_matrix(currpool.lhs, currpool.rhs, e_weights)
assert torch.allclose(result_weights, torch.tensor([[-100.0],
[3.0],
[-100.0]]))
def test_zero_potentials():
currpool = CurrentElems([[torch.tensor(1), 0, 5], [torch.tensor(2), 0,5]],
[[torch.tensor(2), 0, 5]])
e_weights = toy_e_weights_type()
potentials = torch.tensor([0.0,0.0,0.0,0.0,0.0])
desired_match = torch.tensor([[1.0],
[0.0]])
result_match, e_weights = compute_matching(currpool, potentials, e_weights)
assert torch.allclose(result_match, desired_match, atol=1e-6)
def dont_test_tiebreak():
# this is a failing test that reveals a fractional matching
currpool = CurrentElems([[torch.tensor(1), 0, 5],[torch.tensor(1), 0, 5], [torch.tensor(1), 0, 5]],
[[torch.tensor(1), 0, 5], [torch.tensor(1), 0, 5]])
e_weights = toy_e_weights_type()
e_weights_full = weight_matrix(currpool.lhs, currpool.rhs, e_weights)
desired_match = torch.tensor([[1.0,0.0],
[0.0,1.0],
[0.0,0.0]])
result_match, e_weights = compute_matching(currpool, torch.zeros(5), e_weights)
assert torch.allclose(result_match, desired_match, atol=1e-6)
