My understanding is that Unet is a kind of architecture that output an image from an image, and a GAN is the process of having 2 networks that go back and forth together : one to generate images (like Unet), and the other to discrimate those images (are they generated or real ?)
I’m a bit lost on the lesson7-human-numbers section: http://nbviewer.jupyter.org/github/fastai/course-v3/blob/master/nbs/dl1/lesson7-human-numbers.ipynb - what is the goal with this model? What are we trying to achieve?
Word prediction
This s a language model, so we’re trying to predict what the next word is, having read all the previous ones.
We are predicting the next number in a sequence where numbers are represented by words.
This is correct. A unet outputs an image from an image, and a GAN has a generator and a discriminator that go back and forth. In the superres gan notebook Jeremy showed, the generator for the GAN was a unet.
3 Likes
I need to point out- one of the pieces of advice there may be outdated. It says " 11: Dont balance loss via statistics (unless you have a good reason to)". We’re actually doing that exact thing with this line:
switcher = partial(AdaptiveGANSwitcher, critic_thresh=0.65)
That switcher is using a simple heuristic- keep training the critic until it’s firmly “ahead” of the generator by making sure its loss is sufficiently low. This is all pretty new (like…last two weeks) but it appears this is helpful for pretrained networks in particular because it helps to bring the critic in balance with the generator (important- because you don’t want a critic not knowing what it’s doing!)
9 Likes
For the orange ones, the idea is that going from the nth word to the (n+1)th word is the same (regardless of n).
so we use Unet for GAN ?? bu our objective is to make image to look as good like hi res image by resolving the crapiness in the image
but shudnt we learn something new from the each new sequence of word embedding compared to previous ones!! by having same weights, arent we forcing to learn same features from every set of word embedding!? or am i missing something?
I think the main problem I’m having here is that predicting the next word against this data set of english words for numbers is a bit dull - I’d find this section a lot more compelling if it was working against a more interesting word corpus (like multi-word movie names or something)
thank you! translation sounds great idea.
why Gan not on NLP though?
Considering paragraphs of text, there isn’t something distinctive to being the 35th word (as opposed to the 34th word) that you need a network to learn in general.
I’m not sure what you mean by “set of word embedding”? Typically you would use the same embeddings for the whole task.
That’s just for the lesson. We have been doing the exact same thing on IMDB reviews.
Yes, this is a simple task, in the interest of time (since we had only a fraction of the class to cover it in). The super-res was the star task for today.
2 Likes
Yeah, I was having the same issue. This was due to my environment being made months ago before pytorch v1 was released a few weeks ago. During this lecture, I updated my graphics drivers, installed the new pytorch v1 with cuda10 and made a new conda env, new repo, etc and everything is working fine now with 1.0.37.
2 Likes
What’s the best forum section to interact with peers from now on ? This one, or the general part 1 section (or another one) ?
can we collect all interesting blog posts (understandable, well written) by fastai students in a list please?
Can we have Jeremy talk about anything every week. I’ve kinda got used to this.
18 Likes
What a great way to spend our time. Thanks for the teaching! Love what we are going to be able to do with fast.ai!
3 Likes