There are no plans to redo the NLP course at this time, but it is less than a year old and all freely available online: https://www.fast.ai/2019/07/08/fastai-nlp/
why does the loss function need to be negative ?
1 vote to redo NLP course soon
Thanks for sharing!!
Can you explain how the likelihood comes into play? I always thought likelihoods were related to fitting distributions
modification->multiplication
It does not need to be (quite the opposite, it needs to be great when the model performs badly). The log of things between 0 and 1 is negative, so we take the opposite to make it positive.
why does PyTorch have 2 version of the same thing - like a class nn.
and a functional F.
version ? Historical reason ?
I think nn.
is a class and F.
is a function
You can hook into the classes. This allows you to run code based on a layer’s forward or backward pass.
Just to clarify, we have covered notebooks 01, 02, 04 and part of 05. We didn’t cover notebook 03 (ethics) which I am assuming it will be cover next week (based on what Jeremy just mentioned).
It was covered last week.
got it fixed should we write the change? or?
Last week is next week, shhhhh
Jeremy talking about #Masks4All, love this. So much good work here. I will get back to talking on the Discord thread with some of the organizers there to help out with the lobbying effort.
There is a comprehensive answer on the Pytorch forums. The short of it is that nn
modules have states and attributes, whereas F
functionals simply perform operations.
hey @mario_carrillo, 3rd notebook was covered last week, you will find lecture and the ethics notebook cover more or less the same thing.
stay safe
A few of us are also daring each other on twitter to show off the home-made masks using #Masks4allchallenge, I invite you to take the challenge too
Maybe https://discuss.pytorch.org/t/what-is-the-difference-between-torch-nn-and-torch-nn-functional/33597
I think that is actually part of the book and it should be kept as it is.
It is just to show what can go wrong if you don’t Resize
Thank you another great class!