Closed Carayolj closed 1 year ago
In our system, all trainerflows are registered in SUPPORTED_FLOWS. The trainerflow can be either determined by model_name or task_name. If the model can fit in node_classification or link_prediction, which are two major tasks in gnn, it's ok not to create a new flow for this model. Otherwise, we create a new trainerflow for this model and specify the mapping in specific_trainerflow. Since your flow is not specific to a model, I will assume that the flow and the task match each other and you are going to apply existing models to a new scenario. I suggest that you make them identical. Another more flexible option is that you use openhgnn as a package and only import the model you need in your code.
I see, thanks for your answer
Hi, I am trying to create a new trainer flow, as well as a new task. I am struggling a bit and have a few questions: When I register them with
@register_flow(str_flow)
and@register_task(str_task)
, muststr_task
andstr_flow
be identical?Because as my flow is not specific to a model, it is not in the
specific_trainerflow
dictionnary defined in theExperiment
class. So the line 92 in experiment.py(trainerflow = self.specific_trainerflow.get(self.config.model, self.config.task)
) returns the key of the task as the trainerflow_key. Is this the wanted behavior?Thanks!