Large Language Models and Transformers Overview

Large Language Models and Transformers Overview#

TLDR

  • Transformers are now widely used because they

    • scale extremely well

    • perform well across a range of tasks

  • Transformers were introduced in the paper Attention is all you need

    • The general is not different from small language models

    • The challenge a model that can flexibly learn language patterns in an efficient way

  • The key ideas are

    • attention can be further subcomposed into the queries, keys, and values

    • using softmax for layer activation

    • decoder and encoder stacks

Transformers#

Transformers are getting the most attention because they’re the NN architecture which can be reasonably be trained and produces the a great output. Transformers are able to capture patterns in the data in practice that previous architectures could not. Previous model structures markov chain approach weren’t flexible enough to learn the nuances of the same way neural net can. Previous Neural Network architectures had scaling or gradient problems where during training parameter estimation was taking too long or just failed. Transformers are structured in a manner that allows them to be trained readily and stably. And importantly they are just really good in many applications.

The main ideas Transformer#

This is the original picture of a transformer from the paper Attention is all you need. It seems complicated but really a few core intuitive ideas make the whole thing work which is quite amazing. Transformers and this paper, covered extensively so won’t go into detail, we’ll just cover the highlights and provide references to the best materials.

../_images/EncoderDecoder.png

Fig. 3 The famous image from the Attention Is All You Need paper#

The parts we’re going to highlight are

  • Attention, comprised of a Query Key and Value

  • Decoder and Encoder

  • Softmax transform

Attention, Self Attention, and Positional Encoding#

This is the biggest idea as evidenced by the Attention is all you need[VSP+23] paper name itself. The name attention indicates the core idea, In essence the model figures out what parts of its input it should “care about” and “what it shouldn’t. Think of the following question.

I went to the grocery store and hardware store, walked around and bought nails and bread. Which did I buy where?

As a human what do you pay attention to and, and what do you ignore? I’m sure you can explain it generally, but most of us, if asked to write a generalizable mathematical rule, would struggle.

The paper authors did write one such rule, called attention, and as of now nearly everyone is using it.

The core parts of attention are a

  1. Query

  2. Key

  3. Value

Here’s an analogy from Lih Verma

You login to medium and search for some topic of interest — This is Query
Medium has a database of article Title and hash key of article itself — key and Value
Define some similarity between your query and titles in database — ex Dot product of Query and Key
Extract the hash key with maximum match
Return article(Value) corresponding to has key obtained

The explanation in blog post where a transformer is handmade also provides a great intuition, especially because the QKV matrix weights are set by hand leaving no ambiguity whatsoever in the calculation.

Softmax#

Softmax basically takes a vector of arbitrary length and turns it into a proper probability vector. In math it can look a little scary.

\[ \sigma(x)_j = \frac{e^{x_j}}{\sum_k e^{x_k}} \]

But in practice its quite simple. Here’s softmax implemented when implemented in libraries with broadcasting. Here’s the one liner.

x = [.3, 1.2, .8]
np.exp(x)/sum(np.exp(x))

And here’s the output.

array([0.19575891, 0.48148922, 0.32275187])

Now x has been transformed to a proper probability. In transformers softmax is prominently used at the end. However it’s also used within each self attention head as a normalizer and activation function which you can see in Flax here.

Relu activation vs Softmax

I’ve only skimmed this paper but it seems to offer a intuition on why softmax is a good activation function. https://arxiv.org/pdf/2302.06461.pdf

Decoder and Encoder Stacks#

When in the weeds of transformers architecture you’ll hear about encoder models, decoder models, and encoder decoder models. I have a loose grasp on this right now. This website explains it well but there are some nuances between decoder only, and encoder/decoder models that I don’t fully understand.

References#