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

Fix binominal distribution #1860

Merged
merged 5 commits into from
Sep 10, 2024
Merged

Fix binominal distribution #1860

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
0dabbdd
Comments added why the binominal_dispatch function will run into an i…
InfinityMod Sep 9, 2024
baf4d35
Fix for the infinite loop problem of binominal_dispatch.
InfinityMod Sep 9, 2024
1ce6995
Update of the fix to a more concise version.
InfinityMod Sep 10, 2024
ddc8a7c
Linting fix
InfinityMod Sep 10, 2024
ccb4c80
Changed to the proposed solution, to correct log1_p value correctly
InfinityMod Sep 10, 2024
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
2 changes: 2 additions & 0 deletions numpyro/distributions/util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,6 +149,8 @@ def _binom_inv_cond_fn(val):
return cond_exclude_large_mu & (geom_acc <= n)

log1_p = jnp.log1p(-p)
# Make sure p=0 is never taken into account as a fix for possible zeros in p.
log1_p = jnp.where(log1_p == 0, -jnp.finfo(log1_p.dtype).tiny, log1_p)
ret = lax.while_loop(_binom_inv_cond_fn, _binom_inv_body_fn, (-1, key, 0.0))
return ret[0]

Expand Down
Loading