I have enclosed the following error description and two images of the model and guide function. I am not understanding what the key error means.
Just to give a background of what I am doing. I have my prior weights drawn from a categorical distribution and my variational posterior is also categorical
i am passing these sampled weights to a neural network for bayesian inference.
runfile(‘C:/Users/SRIKANTH R/.spyder-py3/pyro_bayesian_ternary.py’, wdir=‘C:/Users/SRIKANTH R/.spyder-py3’)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py”, line 705, in runfile
execfile(filename, namespace)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py”, line 102, in execfile
exec(compile(f.read(), filename, ‘exec’), namespace)
File “C:/Users/SRIKANTH R/.spyder-py3/pyro_bayesian_ternary.py”, line 183, in
instance.do_inference()
File “C:/Users/SRIKANTH R/.spyder-py3/pyro_bayesian_ternary.py”, line 156, in do_inference
elbo=svi.step(images,labels)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\pyro\infer\svi.py”, line 99, in step
loss = self.loss_and_grads(self.model, self.guide, *args, **kwargs)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\pyro\infer\trace_elbo.py”, line 126, in loss_and_grads
loss_particle, surrogate_loss_particle = self._differentiable_loss_particle(model_trace, guide_trace)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\pyro\infer\trace_elbo.py”, line 95, in _differentiable_loss_particle
log_r = _compute_log_r(model_trace, guide_trace)
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\pyro\infer\trace_elbo.py”, line 21, in _compute_log_r
log_r_term = log_r_term - guide_trace.nodes[name][“log_prob”]
File “C:\Users\SRIKANTH R\Anaconda3\lib\site-packages\networkx\classes\reportviews.py”, line 178, in getitem
return self._nodes[n]
KeyError: ‘cn1_prior_dist_sample’