"The value argument must be within the support ... - Pyro Discussion Forum
Can anyone help me with some resources to learn probabilistic programming with pyro? I do not have any background on probabilistic programming. I skimmed through the tutorials and examples in pyro.ai but it seems they assume a background in probabilistic programming concepts.
Ugh, seems like I usually figure out the answer to my question right after caving and posting to a forum about it. You have to provide an arg: ‘optimizer’ to the pyro.optim.PyroLRScheduler(). This optimizer needs to be a class of torch.optim.Optimizer. But it seems that providing a PyroOptim class isn’t allowed. This problem was fixed like so: AdamArgs = { 'lr': 1e-3 } optimizer = torch ...
pyro.sample("y", y_dist, obs=y) # or pyro.factor("y_loss", y_loss_tensor) return ... I am not familiar enough with the details of your model to fix the code for you, but this is covered in detail in the tutorials I linked to, especially the Bayesian regression with SVI section of the Bayesian regression tutorial.
I am using pyro.optim.ClippedAdam and have tried clip_norm = 0.00001, 1.0, 10 and a bunch of values in between, but the gradients are always regardless and don’t appear to change when I change the clip_norm.
I’d like to have the von mises fisher distribution in pyro, so I can score and sample it. Sampling from that distribution is implemented in the Python package geomstats: geomstats/hypersphere.py at master geomstats/geomstats GitHub I am doing amortized inference, and I want to have a deep net that learns the mu (loc) and kappa (concentration) from data, so I need to use pyro.sample(von ...
Von mises fisher 4D - scoring and sampling - Pyro Discussion Forum
Hi. I’m trying to understand hierarchical models, and their Pyro implementation, starting with the eight schools example in the github repository (the SVI one). At first I was confused with the formulation of the model,…