Loading saved models


(Eduardo) #1

Hi. On Lesson 1 we learned how to save our models using:

learn.save('224_dog-cats_lastlayer')

But it’s not very clear how to create a Learn object to load that saved data on it. On Leson 1’s notebook I have, for example, this code to create new “learn” objects:

learn = ConvLearner.pretrained(arch, data, precompute=True, tmp_name=TMP_PATH, models_name=MODEL_PATH)

When I run this -or any other code used on lesson 1 to create “learn” objects- it will start to train the model and that is not what I want, it takes too much time.

How am I supossed to create a new model or learn object and load the save data?

Thanks!


(Elliott Miller) #2

I am wondering this too.


#3

If you have saved a model and want to load it, you would create a new learn object as you describe, then call learn.load('224_dog-cats_lastlayer'). This will populate all the parameters of the learn object with your saved model.