My guide keeps producing nan values, what am I doing wrong?

Misc.
2

Hi @bshabash,
I haven’t fully understood your model, but one quick observation is that you should constraint your pyro.param statements when they are intended to be positive, e.g.

from torch.distributions import constraints

sigma_intercept = pyro.param("sigma_intercept", torch.tensor(1.0),
                             constraint=constraints.positive)
...
    sigma_coef = pyro.param(f"sigma_{i}", torch.tensor(1.0),
                            constraint=constraints.positive)
...
scale_rate = pyro.param("scale_rate", torch.tensor(10.0),
                        constraint=constraints.positive)

Let me know if you still get NAN’s after that fix, and I’ll take a closer look at the censoring logic.