search

isaacmg / model_agnostic_prediction

Peelout model agnostic is a class for easily performing production pipeline predictions.
8 stars 2 forks source link

Graphpipe integration #12

Closed isaacmg closed 6 years ago

isaacmg commented 6 years ago

Graphpipe allows easy serving and exporting of models see https://github.com/oracle/graphpipe-tf-py/blob/master/examples/RemoteModelWithGraphPipe.ipynb . However, it does not address the preprocessing and post processing stages only the actual final steps. Therefore we should include compatibility in model.predict 

from graphpipe import remote
# Predict function for graph pipe
class GraphPipeModel(ModelAgnostic):
   def predict(self, preprocessed_data):
       y = remote.execute(self.graph_pipe_url, preprocessed)
       # Then call postprocess
isaacmg commented 6 years ago

Done. 8c6d3cd83b3c41c50f0609c90eaf233ab8eb088f