Hi!
First, thanks for your time and attention. Second, I am following the DMM model Deep Markov Model — Pyro Tutorials 1.8.6 documentation to build my own model and I am struggling to understand how can I visualize the gradients.
I know in the Gaussian Mixture Model Gaussian Mixture Model — Pyro Tutorials 1.8.6 documentation is easy to do:
# Register hooks to monitor gradient norms. gradient_norms = defaultdict(list) for name, value in pyro.get_param_store().named_parameters(): value.register_hook(lambda g, name=name: gradient_norms[name].append(g.norm().item()))
However in the DMM is not that direct. I need to initialize the DMM somehow, because right now I am obtaining an empty graph
Best regards and thanks for any help