Skip to content

Fine Tuning or Training Certain Layers Exclusively

Domenic Curro edited this page Apr 7, 2016 · 1 revision

Fine-Tuning

Fine-Tuning is the process of training specific sections of a network to improve results.

Making Layers Not Learn

To stop a layer from learning further, you can set it's param attributes in your prototxt.

For example:

layer {
  name: "example"
  type: "example" 
  ...
  param {
    lr_mult: 0    #learning rate of weights
    decay_mult: 1
  }
  param {
    lr_mult: 0    #learning rate of bias
    decay_mult: 0
  }
}