[bug] Unable to render the model in Dirichlet Process Mixture Models in Pyro

[dilaragokay]

Issue Description

I tried to render the model in https://pyro.ai/examples/dirichlet_process_mixture.html using pyro.render_model(model, model_args=(data)). However, I got the following error

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
/tmp/ipykernel_29904/3865686802.py in <module>
     38         pyro.sample("obs", MultivariateNormal(mu[z], torch.eye(2)), obs=data)
     39 
---> 40 pyro.render_model(model, model_args=(data))

~/thesis/pyro/pyro/infer/inspect.py in render_model(model, model_args, model_kwargs, filename, render_distributions)
    492     :rtype: graphviz.Digraph
    493     """
--> 494     relations = get_model_relations(model, model_args, model_kwargs)
    495     graph_spec = generate_graph_specification(relations)
    496     graph = render_graph(graph_spec, render_distributions=render_distributions)

~/thesis/pyro/pyro/infer/inspect.py in get_model_relations(model, model_args, model_kwargs)
    274     with torch.random.fork_rng(), torch.no_grad(), pyro.validation_enabled(False):
    275         with TrackProvenance():
--> 276             trace = poutine.trace(model).get_trace(*model_args, **model_kwargs)
    277 
    278     sample_sample = {}

~/thesis/pyro/pyro/poutine/trace_messenger.py in get_trace(self, *args, **kwargs)
    196         Calls this poutine and returns its trace instead of the function's return value.
    197         """
--> 198         self(*args, **kwargs)
    199         return self.msngr.get_trace()

~/thesis/pyro/pyro/poutine/trace_messenger.py in __call__(self, *args, **kwargs)
    172             )
    173             try:
--> 174                 ret = self.fn(*args, **kwargs)
    175             except (ValueError, RuntimeError) as e:
    176                 exc_type, exc_value, traceback = sys.exc_info()

TypeError: model() takes 1 positional argument but 200 were given

Environment

• OS: Ubuntu 18.04.3 LTS
• Python version: 3.7.11
• PyTorch version: 1.11.0
• Pyro version: 1.8.0+5ab7da2a

Code Snippet

import matplotlib.pyplot as plt
import numpy as np

import torch
import torch.nn.functional as F

import pyro
from pyro.distributions import *
from pyro.infer import SVI
from pyro.optim import Adam

assert pyro.__version__.startswith('1.8.0')
pyro.set_rng_seed(0)

data = torch.cat((MultivariateNormal(-8 * torch.ones(2), torch.eye(2)).sample([50]),
                  MultivariateNormal(8 * torch.ones(2), torch.eye(2)).sample([50]),
                  MultivariateNormal(torch.tensor([1.5, 2]), torch.eye(2)).sample([50]),
                  MultivariateNormal(torch.tensor([-0.5, 1]), torch.eye(2)).sample([50])))

N = data.shape[0]

def mix_weights(beta):
    beta1m_cumprod = (1 - beta).cumprod(-1)
    return F.pad(beta, (0, 1), value=1) * F.pad(beta1m_cumprod, (1, 0), value=1)

def model(data):
    with pyro.plate("beta_plate", T-1):
        beta = pyro.sample("beta", Beta(1, alpha))

    with pyro.plate("mu_plate", T):
        mu = pyro.sample("mu", MultivariateNormal(torch.zeros(2), 5 * torch.eye(2)))

    with pyro.plate("data", N):
        z = pyro.sample("z", Categorical(mix_weights(beta)))
        pyro.sample("obs", MultivariateNormal(mu[z], torch.eye(2)), obs=data)

pyro.render_model(model, model_args=(data))

[ordabayevy]

Can you try this:

-pyro.render_model(model, model_args=(data))
+pyro.render_model(model, model_args=(data,))

I think pyro.render_model expects model_args to be a tuple.

[dilaragokay]

Can you try this:

-pyro.render_model(model, model_args=(data))
+pyro.render_model(model, model_args=(data,))

I think pyro.render_model expects model_args to be a tuple.

That worked, thanks!

[fritzo]

Gosh we should probably assert isinstance(model_args, tuple)