transformer architecture
TRANS-for-mur AR-kuh-tek-chur
ˈtrænsˌfɔːrmər ˈɑːrkɪˌtɛktʃərListen
Hear it in a clear American voice. Slow it down to catch each syllable.
Syllable breakdown
Dictionary meaning
the neural network design introduced in the 2017 paper Attention Is All You Need by Vaswani et al. at Google, which replaced recurrent networks with a self-attention mechanism that computes relationships between all tokens in a sequence simultaneously; the transformer's ability to process sequences in parallel rather than sequentially made it dramatically more efficient to train on large datasets and enabled the scaling laws that produced modern large language models; virtually all state-of-the-art natural language processing, image generation, protein structure prediction, and multimodal AI systems now use transformer-based architectures
Learn one new pronunciation a day
Get a word, name or idiom like “transformer architecture” in your inbox — daily or weekly, free. Unsubscribe any time.