import pytest
from itertools import product
from functools import partial
import torch
from torch.autograd import gradcheck
from entmax.root_finding import sparsemax_bisect, entmax_bisect
from entmax.activations import sparsemax, entmax15
# @pytest.mark.parametrize("dim", (0, 1, 2))
# def test_dim(dim, Map):
# for _ in range(10):
# x = torch.randn(5, 6, 7, requires_grad=True, dtype=torch.float64)
# # gradcheck(f, (x,))
def test_sparsemax():
x = 0.5 * torch.randn(4, 6, dtype=torch.float32)
p1 = sparsemax(x, 1)
p2 = sparsemax_bisect(x)
assert torch.sum((p1 - p2) ** 2) < 1e-7
def test_entmax15():
x = 0.5 * torch.randn(4, 6, dtype=torch.float32)
p1 = entmax15(x, 1)
p2 = entmax_bisect(x, alpha=1.5)
assert torch.sum((p1 - p2) ** 2) < 1e-7
def test_sparsemax_grad():
x = torch.randn(4, 6, dtype=torch.float64, requires_grad=True)
gradcheck(sparsemax_bisect, (x,), eps=1e-5)
@pytest.mark.parametrize("alpha", (1.2, 1.5, 1.75, 2.25))
def test_entmax_grad(alpha):
alpha = torch.tensor(alpha, dtype=torch.float64, requires_grad=True)
x = torch.randn(4, 6, dtype=torch.float64, requires_grad=True)
gradcheck(entmax_bisect, (x, alpha), eps=1e-5)
def test_entmax_correct_multiple_alphas():
n = 4
x = torch.randn(n, 6, dtype=torch.float64, requires_grad=True)
alpha = 1.05 + torch.rand((n, 1), dtype=torch.float64, requires_grad=True)
p1 = entmax_bisect(x, alpha)
p2_ = [
entmax_bisect(x[i].unsqueeze(0), alpha[i].item()).squeeze()
for i in range(n)
]
p2 = torch.stack(p2_)
assert torch.allclose(p1, p2)
def test_entmax_grad_multiple_alphas():
n = 4
x = torch.randn(n, 6, dtype=torch.float64, requires_grad=True)
alpha = 1.05 + torch.rand((n, 1), dtype=torch.float64, requires_grad=True)
gradcheck(entmax_bisect, (x, alpha), eps=1e-5)
@pytest.mark.parametrize("dim", (0, 1, 2, 3))
def test_arbitrary_dimension(dim):
shape = [3, 4, 2, 5]
X = torch.randn(*shape, dtype=torch.float64)
alpha_shape = shape
alpha_shape[dim] = 1
alphas = 1.05 + torch.rand(alpha_shape, dtype=torch.float64)
P = entmax_bisect(X, alpha=alphas, dim=dim)
ranges = [
list(range(k)) if i != dim else [slice(None)]
for i, k in enumerate(shape)
]
for ix in product(*ranges):
x = X[ix].unsqueeze(0)
alpha = alphas[ix].item()
p_true = entmax_bisect(x, alpha=alpha, dim=-1)
assert torch.allclose(P[ix], p_true)
@pytest.mark.parametrize("dim", (0, 1, 2, 3))
def test_arbitrary_dimension_grad(dim):
shape = [3, 4, 2, 5]
alpha_shape = shape
alpha_shape[dim] = 1
f = partial(entmax_bisect, dim=dim)
X = torch.randn(*shape, dtype=torch.float64, requires_grad=True)
alphas = 1.05 + torch.rand(
alpha_shape, dtype=torch.float64, requires_grad=True
)
gradcheck(f, (X, alphas), eps=1e-5)
