What are transformer models and how do they work? • Hacker News Board

Re: What are transformer models and how do they work?

#111

Post by jaidhyani » Sun, Apr 16, 2023, 4:44 PM UTC

andai wrote: post #13

antimora wrote: post #11

Earlier quoted context omitted.

I am trying to learn more in depth. Could you suggest some good resource for learning transformers?

Here's the original paper: https://arxiv.org/abs/1706.03762

The original paper is very good but I would argue it's not well optimized for pedagogy. Among other things, it's targeting a very specific application (translation) and in doing so adopts a more complicated architecture than most cutting-edge modes actually use (encoder-decoder instead of just one or the other). The writers of the paper probably didn't realize they were writing a foundational document at the time. It's good for understanding how certain conventions developed and important historically - but as someone who did read it as an intro to transformers, in retrospect I would have gone with other resources (e.g. "The Illustrated Transformer").

Re: What are transformer models and how do they work?

#112

Post by jaidhyani » Sun, Apr 16, 2023, 4:48 PM UTC

riffraff wrote: post #30

jaidhyani wrote: post #9

Earlier quoted context omitted.

The way the article presents this is misleading. The attention mechanism builds a new vector as a linear combination of other vectors, but after the first layer these have also all been altered by passing through a transformer layer so it makes less sense to talk about "other tokens" in most cases (it becomes increasingly inaccurate the deeper into the model you go). It's also not really moving closer so much as addi…

Isn't "adding" the same as "moving closer" ? E.g. the vector for "bank" is mid-way between the geographical and financial meaning, "bank + money" is closer while "bank + river" if further away.

It depends on the values of the vectors. (4, 4) + (3, 3) results in a new vector (7, 7) which is further away from both contributing vectors than either one was to each other originally. Additionally, negative coefficients are a thing.

Re: What are transformer models and how do they work?

#113

Post by jaidhyani » Sun, Apr 16, 2023, 4:59 PM UTC

joaogui1 wrote: post #66

jaidhyani wrote: post #9

Earlier quoted context omitted.

The way the article presents this is misleading. The attention mechanism builds a new vector as a linear combination of other vectors, but after the first layer these have also all been altered by passing through a transformer layer so it makes less sense to talk about "other tokens" in most cases (it becomes increasingly inaccurate the deeper into the model you go). It's also not really moving closer so much as addi…

You still have one vector per token, that's what they meant, also the fact that the vector associated with each token will ultimately be used to predict the next token, once again showing that it makes sense to talk about other tokens even though they're being transformed inside the model.

Prediction happens at the very end (sometimes functionally earlier, but not always) - most of what happens in the model can be thought of as collecting information in vectors-derived-from-token-embeddings, performing operations on those vectors, and then repeating this process a bunch of times until at some point it results in a meaningful token prediction.

It's pedagogically unfortunate that the residual stream is in the same space as the token embeddings, because it obscures how the residual stream is used as a kind of general compressed-information conduit through the model that attention heads read and write different information to to enable the eventual prediction task.

Re: What are transformer models and how do they work?

#114

Post by jaidhyani » Sun, Apr 16, 2023, 5:05 PM UTC

oergiR wrote: post #38

jaidhyani wrote: post #5

Skimming it, there are a few things about this explanation that rub me just slightly the wrong way. 1. Calling the input token sequence a "command". It probably only makes sense to think of this as a "command" on a model that's been fine-tuned to treat it as such. 2. Skipping over BPE as part of tokenization - but almost every transformer explainer does this, I guess. 3. Describing transformers as using a "word embed…

I agree except for (6). A language model assigns probabilities to sequences. The model needs normalised distributions, eg using a softmax, so that’s the right way of thinking about it.

This is true in general but not in the use case they presented. If they had explained why a normalized distribution is useful it would have made sense - but they just describe this as pick-the-top-answer next-word predictor, which makes the softmax superfluous.