I noticed that the poutine.block is very important for many models (e.g. the code below) by checking all the tutorial articles. But it is still not clear, when should we use it and how?
# Gaussian Mixture Model K = 2 # Fixed number of components. @config_enumerate def model(data): # Global variables. weights = pyro.sample('weights', dist.Dirichlet(0.5 * torch.ones(K))) scale = pyro.sample('scale', dist.LogNormal(0., 2.)) with pyro.plate('components', K): locs = pyro.sample('locs', dist.Normal(0., 10.)) with pyro.plate('data', len(data)): # Local variables. assignment = pyro.sample('assignment', dist.Categorical(weights)) pyro.sample('obs', dist.Normal(locs[assignment], scale), obs=data) global_guide = AutoDelta(poutine.block(model, expose=['weights', 'locs', 'scale']))
The API document of poutine.block is too technical and not friendly for users who just want to focus on building their own model straightforwardly.
Is there a better and easy-understand way to understand the use of poutine.block?