Fixed error

Thought I found a bug, but resolved the problem.
H. Kramer

What are the shapes of your input for the fruit dataset? What are the shapes of the input from the Caltech dataset? Do you resize the images to ensure that the input dimension to the network is consistent? Could you post a sample of your DataLoader code and your training loop?

I did find an error in my code when I checked against the minc-2500 transfer training. Still not sure why it didn’t work in the fruits net the were both resized to 224 but it is working now I think it might have had something to do with the way I initialized the trainer after changing the output layer.