Skip to yearly menu bar Skip to main content


Brainformers: Trading Simplicity for Efficiency

Yanqi Zhou · Nan Du · Yanping Huang · Daiyi Peng · Chang Lan · Da Huang · Siamak Shakeri · David So · Andrew Dai · Yifeng Lu · Zhifeng Chen · Quoc Le · Claire Cui · James Laudon · Jeff Dean

Exhibit Hall 1 #719
[ ]
[ PDF [ Poster


Transformers are central to recent successes in natural language processing and computer vision. Transformers have a mostly uniform backbone where layers alternate between feed-forward and self-attention in order to build a deep network. Here we investigate this design choice and find that more complex blocks that have different permutations of layer primitives can be more efficient. Using this insight, we develop a complex block, named Brainformer, that consists of a diverse sets of layers such as sparsely gated feed-forward layers, dense feed-forward layers, attention layers, and various forms of layer normalization and activation functions. Brainformer consistently outperforms the state-of-the-art dense and sparse Transformers, in terms of both quality and efficiency. A Brainformer model with 8 billion activated parameters per token demonstrates 2x faster training convergence and 5x faster step time compared to its GLaM counterpart. In downstream task evaluation, Brainformer also demonstrates a 3% higher SuperGLUE score with fine-tuning compared to GLaM with a similar number of activated parameters. Finally, Brainformer largely outperforms a Primer dense model derived with NAS with similar computation per token on fewshot evaluations.

Chat is not available.