Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at https://play.google.com/log?format=json&hasfast=true&authuser=0. (Reason: CORS request did not succeed).[Learn More]
This is when I copy/paste into the JavaScript console on Firefox?
why 3e-5 and not 10e -5?
What is fit_one_cycle? It confusing because it does 2 epochs. Appreciate the answer or link to maybe previous lesson if I missed the explanation there
Prints error_rate for each epoch which is 1-accuracy.
When I git update to get the new notebooks, will my notes in old notebook be erased??
Cause it’s between 10e-5 and 10e-4
That’s why we told you to make a duplicate of them to play around 
Is there a way to still get decent results if we can’t find a lot of images for a dataset? (e.g. 20 images per class)
how do you ensure you aren’t just blindly deleting images that your model genuinely failed to predict vs. the intention of deleting actually mislabeled/bad pictures?
Can we use FileDeleter to delete wrong images from training data set? Jeremy showed us how to delete for validation set.
More likes will have the answer from Jeremy
Yes. Create and work on copies of notebooks. E.g. lesson1-copy1.ipynb
What is the criteria - for how many images will be deleted? %age
I agree. It doesn’t seem good to delete images based on top losses when we haven’t actually looked at the examples.
Quick question on the learning rate, why it was taken as 3e-5 to 3e-4. Like why 3 is taken? On the x-axis its 0.13 right?
does fastai has support for image preprocessing to make classifier work better with the images which are not good in quality… there are augmentations but there are many like contours,canny ,contrasting etc which could be required …In real world we may not get many good picture specially satellite images
That is exactly why this is not done automatically. You must delete images that do not correspond to the category, not delete images that are misclassified. In this example, if we have a Google Image of a music band called ‘Grizzly’ we should delete it. If we have an image of a small grizzly that the model predicted as a teddy bear, we should not delete it.
Jeremy is talking about production inference, which I also find fast on CPU, but how come there are these companies coming out with new hardware for inferencing (start ups if I recall)?
Do i need training data during inference?