Adding new classes to an existing model

[marcfielding1]

Hey,

So i'm relatively new to DIGITS we're using via recommendation from NVIDIA - I have a basic question having been through a load of DLI courses and trawled the internet I can't seem to find an answer for:

I have a model trained on 5 classes and 49,000 images on GoogLeNet.

I've just created a dataset containing another 4 classes 37246 images for training.

I go into digits under models I select the model, clone the job, select the NEW dataset, and then select the previous model at the epoch I want to train from.

I then get

ERROR: Cannot copy param 0 weights from layer 'loss1/classifier'; shape mismatch. Source param shape is 5 1024 (5120); target param shape is 4 1024 (4096). To learn this layer's parameters from scratch rather than copying from a saved net, rename the layer.

I get that the shape is different, I'm guessing I need to adjust the model to accept the new dataset but I'm not sure how?

Sorry for the basic question I've literally trawled everything I can find and I just can seem to find an answer the fits for Digits!

Cheers

Marc