CODEWITHSUNDEEP
coreml
sthuettr
sthuettr

Reputation: 1

CoreML updatable Model with multiple Loss Functions

I have a Neural Network that takes sensor data as input and produces eight binary classification outputs. I want to train that Neural Network with CoreML directly on my device. Therefore I need to make the model updatable and set multiple loss functions (one for each output). When I try to compile that model I get the following error message:

Error Domain=com.apple.CoreML Code=3 "Error reading protobuf spec. validator error: This model has more than one loss layers specified, which is not supported at the moment." UserInfo={NSLocalizedDescription=Error reading protobuf spec. validator error: This model has more than one loss layers specified, which is not supported at the moment.}

Now my question is: Is it possible to define multiple loss functions for an updatable CoreML model? If so, how? Any help would be much appreciated!

Upvotes: 0

Views: 110

Answers (1)

Jeshua Lacock
Jeshua Lacock

Reputation: 6688

I'm afraid the error message is very clear: a model with more than one loss layer is not supported. So the answer at this time is, no, it is not possible to have multiple loss layers per model.

Your error (emphasis is mine):

This model has more than one loss layers specified, which is not supported at the moment."

Upvotes: 0

Related Questions