If we really want to use export_savedmodel(serving_input_receiver_fn=?, export_dir_base=?) using custom estimator, (Ignoring the checkpoints) for the production, what it should be? Here's an example of my scripts:
And adding more line in model_fn for converting from dict to tensor.
This solution saves the model using estimator.export_savedmodel(export_dir_base='./Result/', serving_input_receiver_fn=serving_input_fn()).
If we really want to use
export_savedmodel(serving_input_receiver_fn=?, export_dir_base=?)
using custom estimator, (Ignoring the checkpoints) for the production, what it should be? Here's an example of my scripts:UPDATE:
2
And adding more line in
model_fn
for converting fromdict
totensor
. This solution saves the model usingestimator.export_savedmodel(export_dir_base='./Result/', serving_input_receiver_fn=serving_input_fn())
.