I have noticed we can do this as in Python:
const sequential = tf.sequential()
// add layer 1, 2 3...
// now freeze the layers
sequential.trainable = false
// and reuse it in another model
tf.model({inputs:sequential.inputs, outputs})
Is it possible to do it using sequential only? For example:
const sequential = tf.sequential()
// add layers like
sequential.layers.add({...})
sequential.trainable = false
// add now trainable layers
sequential.layers.add({...})
sequential.layers.add({...})
This is useful if the first sequential layers will use weights loaded from elsewhere.