Closed syt123450 closed 5 years ago
For a dynamically input shape model, for example, inceptionv3 from Keras, must provide input shape for prediction. Use TensorSpace Functional model like:
let model = new TSP.models.Model( modelContainer, {
inputs: [ input ],
outputs: [ predictions ],
predictDataShapes: [ [ 299, 299, 3 ] ]
} );
Add predictDataShapes config in model if model's input is dynamically, for example, inceptionv3 model in keras can accept dynamically input shape, if user want to visualize this model in TensorSpace, must config the predictDataShapes in TensorSpace Model.