Hands-On Deep Learning with Go

Updating our CNN to save the model

We need to add a simple function to our CNN example to ensure the model that gets produced is saved, so it can be managed as an object by Pachyderm. Let's add the following to main.go:

func (m *convnet) savemodel() (err error) {  learnables := m.learnables()  var f io.WriteCloser  if f, err = os.OpenFile("model.bin", os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0644); err != nil {    return  }  defer f.Close()  enc := json.NewEncoder(f)  for _, l := range learnables {    t := l.Value().(*tensor.Dense).Data() // []float32    if err = enc.Encode(t); err != nil {      return    }  }  return nil}