Lession 1: Question related Learning rate

Part 1 (2017)
1

As suggested in video lecture , I am experimenting with Orange and Lemons data - My data as follows

print number of files in each folder

training data: Lemons :71
training data: Oranges:73
validation data: Lemons:20
validation data: dogs:21
test data :32

But unlike what was explained in the video, The Learning rate find method runs for 100 %

The plot learning rate Vs iterations came out like this

image
The plot of loss versus learning rate was blank
image

I have tried different combination of Train -validation-test combination but I am not seeing any change > i am wondering where I am going wrong
fyi
I set the batch size as 10
data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz), bs=10)

2

Try calling the loss Vs learning rate plot with argument 1, like that:

learn.sched.plot(1)

This allows for zooming out to see the whole plot as calling without argument shows the last part of the plot.
If you still don’t see anything try reducing the batch size again and repeat what I mentioned above.
Let us know if it works.

1 Like
3

Hi

Got it thanks… This is how it looks. Changed the batch size to 4 as well

image
image.png885×488 36.9 KB

1 Like
4

Glad that it works now. The shape is a good one!