Fast Transformer

This repo implements Fastformer: Additive Attention Can Be All You Need by Wu et al. in TensorFlow. Fast Transformer is a Transformer variant based on additive attention that can handle long sequences efficiently with linear complexity. Fastformer is much more efficient than many existing Transformer models and can meanwhile achieve comparable or even better long text modeling performance.

Installation

Run the following to install:

pip install fast-transformer

Developing fast-transformer

To install fast-transformer, along with tools you need to develop and test, run the following in your virtualenv:

git clone https://github.com/Rishit-dagli/Fast-Transformer.git
# or clone your own fork

cd fast-transformer
pip install -e .[dev]

Usage

import tensorflow as tf
from fast_transformer import FastTransformer

mask = tf.ones([1, 4096], dtype=tf.bool)
model = FastTransformer(
    num_tokens = 20000,
    dim = 512,
    depth = 2,
    max_seq_len = 4096,
    absolute_pos_emb = True, # Absolute positional embeddings
    mask = mask
)
x = tf.experimental.numpy.random.randint(0, 20000, (1, 4096))

logits = model(x) # (1, 4096, 20000)

Want to Contribute ?‍♂️?

Awesome! If you want to contribute to this project, you're always welcome! See Contributing Guidelines. You can also take a look at open issues for getting more information about current or upcoming tasks.

Want to discuss? ?

Have any questions, doubts or want to present your opinions, views? You're always welcome. You can start discussions.

Citation

@misc{wu2021fastformer,
    title   = {Fastformer: Additive Attention is All You Need}, 
    author  = {Chuhan Wu and Fangzhao Wu and Tao Qi and Yongfeng Huang},
    year    = {2021},
    eprint  = {2108.09084},
    archivePrefix = {arXiv},
    primaryClass = {cs.CL}
}

Yannic Kilcher's video was super helpful while building this.