feat: init train v2 #308
feat: init train v2 #308
Conversation
jsolaas
force-pushed
the
feat/train-v2-core
branch
from
ab9bb2a
to
906650a
Compare
jsolaas
force-pushed
the
feat/train-v2-core
branch
from
dc28e33
to
7bdae0f
Compare
| @@ -27,11 +27,11 @@ def add_node(self, node: TNode) -> Self: | |||
| self.nodes[node.id] = node | |||
| return self | |||
|
|
|||
| def add_edge(self, from_id: NodeID, to_id: NodeID) -> Self: | |||
| def add_edge(self, from_id: NodeID, to_id: NodeID, **kwargs) -> Self: | |||
There was a problem hiding this comment.
docstring? intention behind kwargs ... different properties based on type of edge etc?
| @@ -49,6 +49,15 @@ def get_successors(self, node_id: NodeID, recursively=False) -> List[NodeID]: | |||
| else: | |||
| return list(self.graph.successors(node_id)) | |||
|
|
|||
| def get_successor(self, node_id: NodeID) -> NodeID: | |||
There was a problem hiding this comment.
Current limitation that we want to have now, or always, to use "crossovers" when there are more than 1?
There was a problem hiding this comment.
Not always, currently it's used when it's supposed to only be one, but we could of course just use get_successors()[0]. I was thinking we could have both get_successor and get_successors. I think we already have both for either successors or predecessors.
| @@ -58,6 +67,10 @@ def get_predecessor(self, node_id: NodeID) -> NodeID: | |||
| ) | |||
| return predecessors[0] | |||
|
|
|||
| def get_predecessors(self, node_id: NodeID) -> List[NodeID]: | |||
There was a problem hiding this comment.
It's probably easier to see that by looking at where it's used. get_predecessors is used to get all consumers connected to a driver, get_successor is used to get driver for consumer.
Why is this pull request needed?
Initialize the work on a common train, where the features of all current train-models should be handled.