Jan
Jan
reason: for this test, the density estimator has to be off, but it was using too many simulations and became too good. fixed with #1209
@manuelgloeckler has this been fixed by #1151 ?
This is actually solved: in the posterior docstrings we say it's deprecated and if it is passed by a user we issue a deprecation warning. We can leave as it...
link is not working. Should be addressed in #1147
fixed.
Note: `slice_np_vectorized` with a small `thin`, e.g., 1`. Update: thinning does not seem to be the problem. if thinning is small it just increases the likelihood of hitting an inf...
@famura can you please add the new info about best practice regarding `thinning` that you mentioned earlier today? That'd be great! 🙏
> > @famura can you please add the new info about best practice regarding `thinning` that you mentioned earlier today? > > That'd be great! 🙏 > > Sure, where...
TODO: - switch to `slice_np_vectorized` as default arg (in posterior classes and `build_posterior` methods): - [x] switch to `num_chains=10` as default - [x] set `MAF` for SNLE and `NSF` for...
Yes, the `test_process_device` failure is not related to your changes. It seems to be an error in the tests. Can you please tell me the line where it fails (I...