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.

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

Stress test utility for numpyro? #1833

Open
SamuelBrand1 opened this issue Jul 16, 2024 · 2 comments
Open

Stress test utility for numpyro? #1833

SamuelBrand1 opened this issue Jul 16, 2024 · 2 comments
Labels
enhancement New feature or request

Comments

@SamuelBrand1
Copy link

Hi everyone,

I'm working on an ODE based model (solver from diffrax) for respiratory viruses fitted on past seasons of data.

At the moment, we occasionally have maximum iteration fails during NUTS warm-up. This is a bit of a classic problem with ODE models, but I was wondering if there was a stress test utility in numpyro which will sample parameters (in the unconstrained domain I guess) and record the parameters that cause a fail in the model log posterior density call (e.g. like this from LogDensityProblems.jl.

I had a look in the docs but I might well have missed something, apols in advance if I have.

@fehiepsi fehiepsi added enhancement New feature or request question Further information is requested and removed enhancement New feature or request labels Jul 19, 2024
@fehiepsi
Copy link
Member

I think you can draw samples from prior and use the utility log_density to check which of them causes NaN. We don't have a utility for that but please feel free to submit a PR. The feature would be useful for many usage cases.

@fehiepsi fehiepsi added enhancement New feature or request and removed question Further information is requested labels Jul 19, 2024
@SamuelBrand1
Copy link
Author

Cheers for the update.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

2 participants