Concept
Transformer
A Transformer is a neural network architecture, introduced in 2017 ("Attention Is All You Need"), that processes a sequence of tokens by letting every token attend to every other token in parallel. This is the key contrast with the RNNs and LSTMs that preceded it: instead of…
The rest of “Transformer” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→