A basic example with PyroModule

Tutorials
1

Last time I used Pyro was few years back when the PyroModule interface wasn’t there. Now I am back to Pyro again and trying to use it to replicate the coin fairness example. I am following the examples here for understanding the PyroModule interface.

device = torch.device('cpu')
data = torch.rand(1000, device=device)
data[data>0.7] = 1.; data[data<=0.7] = 0.

class Model(PyroModule):
    def __init__(self):
        super().__init__()
        self.fairness = PyroSample(prior=dist.Beta(10., 10.))
    
    def forward(self, data=None):
        with pyro.plate("obs", len(data)):
            return pyro.sample("data", self.fairness, obs=data)

class Guide(PyroModule):
    def __init__(self):
        super().__init__()
        self.alpha = PyroParam(torch.tensor(15.), constraint=dist.constraints.positive)
        self.beta = PyroParam(torch.tensor(15.), constraint=dist.constraints.positive)

        self.fairness = PyroSample(prior=dist.Beta(self.alpha, self.beta))
    
    def forward(self, _):
        return self.fairness

model = Model()
guide = Guide()
svi = infer.SVI(model, guide, optimizer, loss=infer.Trace_ELBO())

But I am pretty sure I am making some stupid mistake and that’s why upon svi.step(data), I am getting this error

Traceback (most recent call last):
  File "coin_fairness.py", line 69, in <module>
    L.append(svi.step(data))
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/infer/svi.py", line 128, in step
    loss = self.loss_and_grads(self.model, self.guide, *args, **kwargs)
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/infer/trace_elbo.py", line 126, in loss_and_grads
    for model_trace, guide_trace in self._get_traces(model, guide, args, kwargs):
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/infer/elbo.py", line 170, in _get_traces
    yield self._get_trace(model, guide, args, kwargs)
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/infer/trace_elbo.py", line 52, in _get_trace
    model_trace, guide_trace = get_importance_trace(
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/infer/enum.py", line 55, in get_importance_trace
    model_trace.compute_log_prob()
  File "/home/ayan/anaconda3/lib/python3.8/site-packages/pyro/poutine/trace_struct.py", line 216, in compute_log_prob
    log_p = site["fn"].log_prob(site["value"], *site["args"], **site["kwargs"])
AttributeError: 'Tensor' object has no attribute 'log_prob'

Can someone point me out what am I doing wrong ?

1 Like
2

Accessing self.fairness in forward calls pyro.sample automatically the first time it is accessed: that is, self.fairness is already a sample, not a distribution.

You will also need to define a likelihood for the data and leave fairness latent in the model, e.g. a Bernoulli distribution if your data are binary:

class Model(PyroModule):
    ...
    def forward(self, data=None):
        fairness = self.fairness. # must be accessed for the first time outside the plate
        with pyro.plate("obs", len(data)):
            return pyro.sample("data", dist.Bernoulli(fairness), obs=data)

See part 1 of the SVI tutorial for more background on the difference between latent variables, observed variables and parameters.

1 Like
3

Ah right, while porting from the function-style model() and guide(), I totally missed the dist.Bernoulli.

Anyway, thanks