Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Make one sided truncated normal distributions differentiable #10

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Make one sided truncated normal distributions differentiable #10

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 10 additions & 2 deletions TruncatedNormal.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,11 +73,19 @@ def auc(self):

@staticmethod
def _little_phi(x):
return (-(x ** 2) * 0.5).exp() * CONST_INV_SQRT_2PI
if x.isinf():
return torch.zeros(x.size()).to(x)
else:
return (-(x ** 2) * 0.5).exp() * CONST_INV_SQRT_2PI

@staticmethod
def _big_phi(x):
return 0.5 * (1 + (x * CONST_INV_SQRT_2).erf())
if x.isposinf():
return torch.ones(x.size()).to(x)
elif x.isneginf():
return torch.zeros(x.size()).to(x)
else:
return 0.5 * (1 + (x * CONST_INV_SQRT_2).erf())

@staticmethod
def _inv_big_phi(x):
Expand Down
27 changes: 25 additions & 2 deletions tests/test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,12 @@ def _test_numerical(self, loc, scale, a, b, do_icdf=True):
N = 10
for i in range(N):
p = i / (N - 1)
x = a + (b - a) * p
if torch.isinf(torch.tensor(a)):
x = b - scale * i / N
elif torch.isinf(torch.tensor(b)):
x = a + scale * i / N
else:
x = a + (b - a) * p

cdf_sc = sc.cdf(x)
cdf_pt = float(pt.cdf(torch.tensor(x)))
Expand All @@ -84,6 +89,18 @@ def _test_numerical(self, loc, scale, a, b, do_icdf=True):
icdf_pt = float(pt.icdf(torch.tensor(p)))
self.assertRelativelyEqual(icdf_sc, icdf_pt, tol=1e-4, error=1e-3)

def _test_grad(self, loc, scale, a, b, grad_point):
loc = torch.nn.parameter.Parameter(torch.tensor(loc))
scale = torch.nn.parameter.Parameter(torch.tensor(scale))
pt = TruncatedNormalPT(loc, scale, a, b)
grads = torch.autograd.grad(pt.log_prob(grad_point), [loc, scale])
self.assertFalse(any([grad.isnan() for grad in grads]))

def test_grad(self):
self._test_grad(0., 1., -2., 0., -1.)
self._test_grad(0., 1., -2., torch.inf, -1.)
self._test_grad(0., 1., -torch.inf, 0., -1.)

def test_simple(self):
self._test_numerical(0., 1., -2., 0.)
self._test_numerical(0., 1., -2., 1.)
Expand All @@ -95,6 +112,10 @@ def test_simple(self):
self._test_numerical(0., 1., 0., 2.)
self._test_numerical(1., 2., 1., 2.)
self._test_numerical(1., 2., 2., 4.)
self._test_numerical(0., 1., -2., torch.inf)
self._test_numerical(0., 1., -torch.inf, 0.)
self._test_numerical(1., 2., 2., torch.inf)
self._test_numerical(1., 2., -torch.inf, 4.)

def test_precision(self):
self._test_numerical(0., 1., 2., 3.)
Expand All @@ -112,7 +133,9 @@ def test_support(self):
pt = TruncatedNormalPT(0., 1., -1., 2., validate_args=None)
with self.assertRaises(ValueError) as e:
pt.log_prob(torch.tensor(-10))
self.assertEqual(str(e.exception), 'The value argument must be within the support')

self.assertFalse(str(e.exception) != 'The value argument must be within the support' and
str(e.exception) != 'Expected value argument (Tensor of shape ()) to be within the support (Interval(lower_bound=-1.0, upper_bound=2.0)) of the distribution TruncatedNormal(a: -1.0, b: 2.0), but found invalid values:\n-10.0')

def test_cuda(self):
if not torch.cuda.is_available():
Expand Down