The two models in the Jazz assignment

Course Q&A Deep Learning Specialization Sequence Models
1

I am confused about the designation of the two models in the Jazz assignment (Week 1 of Course 5) and their usage:

  1. Where in the code is the trained djmodel used?
  2. Why do we need the inference_model? Where in the code do we train this model or read pre-trained parameters for it?
1 Like
2

Hi Meir,

The djmodel is used for training so you use the ground truth lables (i.e. x → Note sequence, Y → Note sequence shifted one note to the left). You use this model in cell 7
model = djmodel(Tx=30, LSTM_cell=LSTM_cell, densor=densor, reshaper=reshaper)

On the other hand, you use your inference model for prediction where the last output is the input for the new step.

Hope this answers your question…

3

That was exactly my question. Namely, djmodel is trained, but is not used for predictions. So, the question arises: “if we don’t use this model to make predictions, then what did we train it for?”

And the opposite question for the inference_model. Namely, how can we use a model for predictions without having trained it first?

4

Hi Meir,

My understanding is that you use the weights learned by djmodel in the inference_model(the LSTM_cell, densor…) maybe someone can confirm…

5

@juancopi81 is right. The models created by two functions are sharing the actual layer object instances.
LSTM_cell and densor are global variables that are passed to both functions. The djmodel() function creates a model that trains these layers. After that model is compiled and fit, the layers have learned to recognize the jazz patterns. These same layer instances are then passed to music_inference_model() function which uses them to predict the music notes.

1 Like