"Attention is all you need" paper digested (2018)

For a higher level shorter overview, I found this video informative: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;SZorAJ4I-sA?si=pnfzZ17PYQfV4aqq" rel="nofollow noreferrer">https:&#x2F;&#x2F;youtu.be&#x2F;SZorAJ4I-sA?si=pnfzZ17PYQfV4aqq</a>

IMO understanding this kind of content depends on where you&#x27;re starting out, or you&#x27;re getting too much or too little explanation of the supporting components.<p>Personally, I used ChatGPT to help me understand this paper from the stance of a senior engineer with some ML experience, and followed up to clarify where I had questions: <a href="https:&#x2F;&#x2F;chat.openai.com&#x2F;share&#x2F;c17dbcfc-7e3a-44fd-aa88-28c11702ed86" rel="nofollow noreferrer">https:&#x2F;&#x2F;chat.openai.com&#x2F;share&#x2F;c17dbcfc-7e3a-44fd-aa88-28c117...</a>

Attention is super helpful when you have enough data but it didn’t really help me on a problem with a tiny dataset

<a href="https:&#x2F;&#x2F;www.askyoutube.ai&#x2F;share&#x2F;6521aa6077733b7c0ad24b55" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.askyoutube.ai&#x2F;share&#x2F;6521aa6077733b7c0ad24b55</a>

I need this breakdown, I tried to read the paper but it&#x27;s not accessible to a more than average interested layman

Google regrets publishing this paper!

Love Yannic

Actually, all you need is all you need—anything that claims otherwise is just wrong. In other words, you need all the things that you need.