Multi-level modeling support from AMRs #614

djinnome · 2024-09-25T20:14:01Z

Ben Gyori has now support for expressions in distribution parameters so we can generate multi-level models. We need to properly sort the expressions so they are evaluated in order.

djinnome · 2024-10-02T20:17:45Z

Code to sample from parameter distributions should probably go here:

pyciemss/pyciemss/mira_integration/compiled_dynamics.py

Line 84 in 08cf9fc

def _compile_param_values_mira(

djinnome · 2024-10-28T23:51:21Z

Hi @SamWitty

I have a question about how to sample from a PyroSample object.

In the _compile_param_values_mira function below, I have topologically sorted the parameters in mira, and now I am compiling each parameter in order:

pyciemss/pyciemss/mira_integration/compiled_dynamics.py

Lines 83 to 110 in e48ff2f

    
           @_compile_param_values.register(mira.modeling.Model) 
        
           def _compile_param_values_mira( 
        
               src: mira.modeling.Model, 
        
           ) -> Dict[str, Union[torch.Tensor, pyro.nn.PyroParam, pyro.nn.PyroSample]]: 
        
               values = {} 
        
               for param_name in sort_mira_dependencies(src): 
        
                   param_info = src.parameters[param_name] 
        
                   #param_name = get_name(param_info) 
        
                   if param_info.placeholder: 
        
                       continue 
        
                   param_dist = getattr(param_info, "distribution", None) 
        
                   if param_dist is None: 
        
                       param_value = param_info.value 
        
                   else: 
        
                       param_value = mira_distribution_to_pyro(param_dist, free_symbols=values) 
        
                   if isinstance(param_value, torch.nn.Parameter): 
        
                       values[param_name] = pyro.nn.PyroParam(param_value) 
        
                   elif isinstance(param_value, pyro.distributions.Distribution): 
        
                       values[param_name] = pyro.nn.PyroSample(param_value) 
        
                   elif isinstance(param_value, (numbers.Number, numpy.ndarray, torch.Tensor)): 
        
                       values[param_name] = torch.as_tensor(param_value, dtype=torch.float32) 
        
                   else: 
        
                       raise TypeError(f"Unknown parameter type: {type(param_value)}") 
        
               return values

In particular, on line 99, I pass the values of the parameters that have already been compiled to Pyro as free symbols that are used to parse the sympy expressions:

pyciemss/pyciemss/mira_integration/compiled_dynamics.py

Line 99 in e48ff2f

param_value = mira_distribution_to_pyro(param_dist, free_symbols=values)

However, to my surprise, the values dictionary that contains the parameter values has not been evaluated yet, so instead of the broadcast_tensors(*tensors) function receiving a list of tensors as its argument, it gets a list containing a PyroSample object instead:

pyciemss/compiled_dynamics.py:26: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:909: in wrapper
    return dispatch(args[0].__class__)(*args, **kw)
pyciemss/mira_integration/compiled_dynamics.py:99: in _compile_param_values_mira
    param_value = mira_distribution_to_pyro(param_dist, free_symbols=values)
pyciemss/mira_integration/distributions.py:419: in mira_distribution_to_pyro
    k: safe_sympytorch_parse_expr(v, local_dict=free_symbols)
pyciemss/mira_integration/distributions.py:53: in safe_sympytorch_parse_expr
    return sympytorch.SymPyModule(expressions=[expr.args[0]])(**local_dict).squeeze()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/torch/nn/modules/module.py:1511: in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/torch/nn/modules/module.py:1520: in _call_impl
    return forward_call(*args, **kwargs)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/sympytorch/sympy_module.py:265: in forward
    out = torch.broadcast_tensors(*out)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

tensors = (PyroSample(prior=Beta()),)

    def broadcast_tensors(*tensors):
        r"""broadcast_tensors(*tensors) -> List of Tensors
    
        Broadcasts the given tensors according to :ref:`broadcasting-semantics`.
    
        Args:
            *tensors: any number of tensors of the same type
    
        .. warning::
    
            More than one element of a broadcasted tensor may refer to a single
            memory location. As a result, in-place operations (especially ones that
            are vectorized) may result in incorrect behavior. If you need to write
            to the tensors, please clone them first.
    
        Example::
    
            >>> x = torch.arange(3).view(1, 3)
            >>> y = torch.arange(2).view(2, 1)
            >>> a, b = torch.broadcast_tensors(x, y)
            >>> a.size()
            torch.Size([2, 3])
            >>> a
            tensor([[0, 1, 2],
                    [0, 1, 2]])
        """
        # This wrapper exists to support variadic args.
        if has_torch_function(tensors):
            return handle_torch_function(broadcast_tensors, tensors, *tensors)
>       return _VF.broadcast_tensors(tensors)  # type: ignore[attr-defined]
E       TypeError: expected Tensor as element 0 in argument 0, but got PyroSample

The reason for this is described in the Pyro documentation:

https://docs.pyro.ai/en/stable/nn.html#pyro.nn.module.PyroSample

assert isinstance(my_module, PyroModule)
my_module.x = PyroSample(Normal(0, 1))                    # independent
my_module.y = PyroSample(lambda self: Normal(self.x, 1))  # dependent

Note that my_module.y will not evaluate the lambda expression in the PyroSample object until getattr or setattr is called. Similarly, in the _compile_param_values_mira function, the pyro.distributions.Distribution object on line 103 is wrapped in a PyroSample object on line 104:

pyciemss/pyciemss/mira_integration/compiled_dynamics.py

Lines 103 to 104 in e48ff2f

    
           elif isinstance(param_value, pyro.distributions.Distribution): 
        
               values[param_name] = pyro.nn.PyroSample(param_value)

If the pyro.distributions.Distribution object contains unevaluated dependencies, such as the dependency of beta_mean on gamma_mean in the example below, then when should this dependency get evaluated?

beta_mean = Parameter(name='beta_mean',
                    distribution=Distribution(type="Beta1",
                    parameters={'alpha': sympy.Integer(10)*sympy.Symbol("gamma_mean"),
                                'beta': sympy.Integer(10)}))
gamma_mean = Parameter(name='gamma_mean',
                    distribution=Distribution(type="InverseGamma1",
                    parameters={'alpha': sympy.Integer(10),
                                'beta': sympy.Integer(10)}))

The unevaluated PyroSample objects are currently being held in the values dictionary, so when should I compile the values dictionary to a dictionary of tensors (let's call it compiled_values) so that I can compile the parameters that depend on their values?

Here are some options:

inside mira_distributions_to_pyro() where mira distribution parameters are passed to their corresponding pyro distribution function:

pyciemss/pyciemss/mira_integration/distributions.py

Lines 418 to 423 in e48ff2f

    
           parameters = { 
        
               k: safe_sympytorch_parse_expr(v, local_dict=free_symbols)  
        
               if isinstance(v, SympyExprStr)  
        
               else torch.as_tensor(v)  
        
               for k, v in mira_dist.parameters.items() 
        
           }

Inside _compile_param_values_mira() where distributions are wrapped in a PyroSample:

pyciemss/pyciemss/mira_integration/compiled_dynamics.py

Lines 103 to 104 in e48ff2f

    
           elif isinstance(param_value, pyro.distributions.Distribution): 
        
               values[param_name] = pyro.nn.PyroSample(param_value)

Inside the pyciemss.compiled_dynamics.CompiledDynamics.__init__() method where samples are evaluated:

pyciemss/pyciemss/compiled_dynamics.py

Lines 36 to 39 in e48ff2f

    
           if isinstance( 
        
               v, (torch.nn.Parameter, pyro.nn.PyroParam, pyro.nn.PyroSample) 
        
           ): 
        
               setattr(self, f"persistent_{get_name(k)}", v)

It seems that I would need to perform options 1-3 on gamma_mean parameter before I could perform options 1-3 on the beta_mean parameter

Any advice would be appreciated.

djinnome · 2025-01-08T15:28:58Z

The current issue is now that there is a RecursionError:

git checkout 614-multi-level-modeling-support-from-amrs  
pytest tests/test_compiled_dynamics.py -k hierarchical

results in:

======================================================================================= FAILURES ========================================================================================
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model0-acyclic_model0] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x324891250>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a921e0>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/beta_mean_cycle_sir_model.json
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model0-acyclic_model1] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x3248b0500>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a921e0>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/beta_mean_cycle_sir_model.json
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model1-acyclic_model0] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x324891250>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a92420>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/beta_mean_gamma_cycle_sir_model.json
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model1-acyclic_model1] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x3248b0500>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a92420>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/beta_mean_gamma_cycle_sir_model.json
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model2-acyclic_model0] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x324891250>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a92390>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/gamma_mean_beta_mean_cycle_sir_model.json
______________________________________________________ test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model2-acyclic_model1] _______________________________________________________

acyclic_model = <tests.fixtures.ModelFixture object at 0x3248b0500>, cyclic_model = <tests.fixtures.ModelFixture object at 0x324a92390>, start_time = 0.0, end_time = 40.0

    @pytest.mark.parametrize("acyclic_model", ACYCLIC_MODELS)
    @pytest.mark.parametrize("cyclic_model", CYCLIC_MODELS)
    @pytest.mark.parametrize("start_time", START_TIMES)
    @pytest.mark.parametrize("end_time", END_TIMES)
    def test_hierarchical_compiled_dynamics(
        acyclic_model, cyclic_model, start_time, end_time
    ):
        """
        Test the loading and dependency analysis of hierarchical compiled dynamics models.
    
        This test verifies the following:
        - An acyclic MIRA model can be loaded from a URL and is of type TemplateModel.
        - The dependencies of the acyclic MIRA model are sorted as expected.
        - Attempting to sort dependencies for a cyclic MIRA model raises a NetworkXUnfeasible exception.
        - A CompiledDynamics model can be loaded from an acyclic URL and is of the correct type.
        - The prior and posterior dependencies of the CompiledDynamics model match the expected structure
          when given the specified start and end times.
        """
        acyclic_mira_model = model_from_url(acyclic_model.url)
        assert isinstance(acyclic_mira_model, mira.metamodel.TemplateModel)
        #assert sort_mira_dependencies(acyclic_mira_model) == acyclic_model.important_parameter
        with pytest.raises(nx.NetworkXUnfeasible):
            cyclic_mira_model = model_from_url(cyclic_model.url)
            print(cyclic_model.url)
            sort_mira_dependencies(cyclic_mira_model)
>       model = CompiledDynamics.load(acyclic_model.url)

tests/test_compiled_dynamics.py:109: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:170: in _load_from_url_or_path
    return cls.load(model)
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:180: in _load_from_template_model
    return cls.load(mira.modeling.Model(template))
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/functools.py:946: in _method
    return method.__get__(obj, cls)(*args, **kwargs)
pyciemss/compiled_dynamics.py:185: in _load_from_mira_model
    return cls(src)
pyciemss/compiled_dynamics.py:66: in __init__
    self.instantiate_parameters()
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:829: in cached_fn
    return fn(self, *args, **kwargs)
pyciemss/compiled_dynamics.py:110: in instantiate_parameters
    param_val = getattr(self, f"persistent_{param_name}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
pyciemss/mira_integration/compiled_dynamics.py:140: in param_value
    k: getattr(model, f"persistent_{k}")
../../../../.pyenv/versions/miniconda3-3.11-24.1.2-0/envs/pyciemss/lib/python3.12/site-packages/pyro/nn/module.py:607: in __getattr__
    prior = prior(self)
E   RecursionError: maximum recursion depth exceeded
!!! Recursion detected (same locals & position)
--------------------------------------------------------------------------------- Captured stdout call ----------------------------------------------------------------------------------
https://raw.githubusercontent.com/DARPA-ASKEM/simulation-integration/main/data/models/gamma_mean_beta_mean_cycle_sir_model.json
================================================================================ short test summary info ================================================================================
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model0-acyclic_model0] - RecursionError: maximum recursion depth exceeded
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model0-acyclic_model1] - RecursionError: maximum recursion depth exceeded
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model1-acyclic_model0] - RecursionError: maximum recursion depth exceeded
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model1-acyclic_model1] - RecursionError: maximum recursion depth exceeded
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model2-acyclic_model0] - RecursionError: maximum recursion depth exceeded
FAILED tests/test_compiled_dynamics.py::test_hierarchical_compiled_dynamics[40.0-0.0-cyclic_model2-acyclic_model1] - RecursionError: maximum recursion depth exceeded
=========================================================================== 6 failed, 33 deselected in 22.28s ========================================================================

djinnome self-assigned this Sep 25, 2024

djinnome linked a pull request Oct 21, 2024 that will close this issue

Added notebook for hierarchical sir model and AMR #620

Draft

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Multi-level modeling support from AMRs #614

Multi-level modeling support from AMRs #614

djinnome commented Sep 25, 2024

djinnome commented Oct 2, 2024

djinnome commented Oct 28, 2024 •

edited

Loading

djinnome commented Jan 8, 2025

Multi-level modeling support from AMRs #614

Multi-level modeling support from AMRs #614

Comments

djinnome commented Sep 25, 2024

djinnome commented Oct 2, 2024

djinnome commented Oct 28, 2024 • edited Loading

djinnome commented Jan 8, 2025

djinnome commented Oct 28, 2024 •

edited

Loading