Request: add user-defined attribute into containers.Graph

[0h-n0]

• inferno version: 0.4.0
• Python version: 3.7
• Operating System: Ubuntu-18.04

Description

I would like to set attributes on nodes of containers.Graph. If the features is implemented, we can use it more Graph-likely.

What I Did(request code)

from inferno.extensions.containers import Graph

module = Graph()
module.add_input_node('input')
module.add_node('l1', nn.Linear(3, 18), previous='input', attributes={"hoge": [1, 2, 3]})

g = module.graph
print(g.nodes(data=True))
# [('input', {'is_input_node': True}), ('l1', {'attributes': {"hoge": [1, 2, 3]}})]
# or
# [('input', {'is_input_node': True}), ('l1', {"hoge": [1, 2, 3]})]

[constantinpape]

If I understand your code snippet correctly, you want to be able to pass a dict attributes to add_node, correct? How would you use these in the nodes? Access them via node.attributes?

[0h-n0]

Thank you for your replay.

If I understand your code snippet correctly, you want to be able to pass a dict attributes to add_node, correct?

Yes, your right.

How would you use these in the nodes? Access them via node.attributes?

I want to use them through native Networkx interfaces. Networkx provides the following interfaces. We can't access them via node.attributes.

>> G.add_node(1, time='5pm')
>>> G.add_nodes_from([3], time='2pm')
>>> G.nodes[1]
{'time': '5pm'}
>>> G.nodes[1]['room'] = 714
>>> G.nodes.data()
NodeDataView({1: {'time': '5pm', 'room': 714}, 3: {'time': '2pm'}})

The following code is simple implementation (Graph class in inferno/extensions/containers/graph.py), I guess.

def add_node(self, name, module, previous=None, **attr):       
        """
        Add a node to the graph.

        Parameters
        ----------  
        name : str   
            Name of the node. Nodes are identified by their names.  
        attr : dict 
            Attributes of the Nodes.    

        module : torch.nn.Module         
            Torch module for this node.                                                

        previous : str or list of str            
            (List of) name(s) of the previous node(s).          
           
        Returns 
        -------   
        Graph    
            self     
        """          
        assert isinstance(module, nn.Module)           
        self.add_module(name, module)                     
        self.graph.add_node(name, **attr)                  
        if previous is not None:                                     
            for _previous in pyu.to_iterable(previous): 
                self.add_edge(_previous, name)  
        return self