English

FIT: Far-reaching Interleaved Transformers

Machine Learning 2023-05-26 v2 Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition

Abstract

We present FIT: a transformer-based architecture with efficient self-attention and adaptive computation. Unlike original transformers, which operate on a single sequence of data tokens, we divide the data tokens into groups, with each group being a shorter sequence of tokens. We employ two types of transformer layers: local layers operate on data tokens within each group, while global layers operate on a smaller set of introduced latent tokens. These layers, comprising the same set of self-attention and feed-forward layers as standard transformers, are interleaved, and cross-attention is used to facilitate information exchange between data and latent tokens within the same group. The attention complexity is O(n2)O(n^2) locally within each group of size nn, but can reach O(L4/3)O(L^{{4}/{3}}) globally for sequence length of LL. The efficiency can be further enhanced by relying more on global layers that perform adaptive computation using a smaller set of latent tokens. FIT is a versatile architecture and can function as an encoder, diffusion decoder, or autoregressive decoder. We provide initial evidence demonstrating its effectiveness in high-resolution image understanding and generation tasks. Notably, FIT exhibits potential in performing end-to-end training on gigabit-scale data, such as 6400×\times6400 images, or 160K tokens (after patch tokenization), within a memory capacity of 16GB, without requiring specific optimizations or model parallelism.

Keywords

Cite

@article{arxiv.2305.12689,
  title  = {FIT: Far-reaching Interleaved Transformers},
  author = {Ting Chen and Lala Li},
  journal= {arXiv preprint arXiv:2305.12689},
  year   = {2023}
}

Comments

preliminary work (code at https://github.com/google-research/pix2seq)

R2 v1 2026-06-28T10:40:52.236Z