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

Avoid sampling -inf from soft Laplace distribution (solves #1186) #1189

Merged
merged 5 commits into from
Oct 13, 2021
+4 −3
Merged

Avoid sampling -inf from soft Laplace distribution (solves #1186) #1189

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7894356
Avoid infinities when doing inverse-transform sampling in truncated d…
Oct 11, 2021
e3eba69
Remove use of jax.config. Do not use maxval in random.uniform as it i…
Oct 12, 2021
8bf3e41
lint and format
Oct 12, 2021
b714aab
Avoid sampling -inf from soft Laplace distribution
Oct 13, 2021
3c052b4
Merge branch 'pyro-ppl:master' into enhancement/softlaplace
omarfsosa Oct 13, 2021
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
7 changes: 4 additions & 3 deletions numpyro/distributions/continuous.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1349,9 +1349,10 @@ def log_prob(self, value):

def sample(self, key, sample_shape=()):
assert is_prng_key(key)
u = random.uniform(
key, shape=sample_shape + self.batch_shape + self.event_shape
)
dtype = jnp.result_type(float)
finfo = jnp.finfo(dtype)
minval = finfo.tiny
u = random.uniform(key, shape=sample_shape + self.batch_shape, minval=minval)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also, thanks for cleaning up the event shape. +1

return self.icdf(u)

# TODO: refactor validate_sample to only does validation check and use it here
Expand Down