Implementation of the Transformer variant proposed in "Transformer Quality in Linear Time"

Related tags

Deep Learningdeep-learningtransformersartificial-intelligenceattention-mechanismefficient-transformers
Overview

FLASH - Pytorch

Implementation of the Transformer variant proposed in the paper Transformer Quality in Linear Time

Install

$ pip install FLASH-pytorch

Usage

The main novel circuit in this paper is the "Gated Attention Unit", which they claim can replace multi-headed attention while reducing it to just one head.

It uses a relu squared activation in place of the softmax, the activation of which was first seen in the Primer paper, and the use of ReLU in ReLA Transformer. The gating style seems mostly inspired by gMLPs.

import torch
from flash_pytorch import GAU

gau = GAU(
    dim = 512,
    query_key_dim = 128,     # query / key dimension
    causal = True,           # autoregressive or not
    expansion_factor = 2,    # hidden dimension = dim * expansion_factor
)

x = torch.randn(1, 1024, 512)
out = gau(x) # (1, 1024, 512)

The authors then combine GAU with Katharopoulos linear attention, using grouping of the sequences to overcome a known issue with autoregressive linear attention.

This combination of the quadratic gated attention unit with grouped linear attention they named FLASH

You can also use this quite easily

import torch
from flash_pytorch import FLASH

flash = FLASH(
    dim = 512,
    group_size = 256,             # group size
    causal = True,                # autoregressive or not
    query_key_dim = 128,          # query / key dimension
    expansion_factor = 2.         # hidden dimension = dim * expansion_factor
)

x = torch.randn(1, 1111, 512)     # sequence will be auto-padded to nearest group size
out = flash(x) # (1, 1111, 512)

Finally, you can use the full FLASH transformer as mentioned in the paper. This contains all the positional embeddings mentioned in the paper. Absolute positional embedding uses scaled sinusoidal. GAU quadratic attention will get one-headed T5 relative positional bias. On top of all this, both GAU attention as well as the linear attention will be rotary embedded (RoPE).

import torch
from flash_pytorch import FLASHTransformer

model = FLASHTransformer(
    num_tokens = 20000,          # number of tokens
    dim = 512,                   # model dimension
    depth = 12,                  # depth
    causal = True,               # autoregressive or not
    group_size = 256,            # size of the groups
    query_key_dim = 128,         # dimension of queries / keys
    expansion_factor = 2.,       # hidden dimension = dim * expansion_factor
    norm_type = 'scalenorm',     # in the paper, they claimed scalenorm led to faster training at no performance hit. the other option is 'layernorm' (also default)
    shift_tokens = True          # discovered by an independent researcher in Shenzhen @BlinkDL, this simply shifts half of the feature space forward one step along the sequence dimension - greatly improved convergence even more in my local experiments
)

x = torch.randint(0, 20000, (1, 1024))
logits = model(x) # (1, 1024, 20000)

Test on Autoregressive Enwik8

$ python train.py

Citations

@article{Hua2022TransformerQI,
    title   = {Transformer Quality in Linear Time},
    author  = {Weizhe Hua and Zihang Dai and Hanxiao Liu and Quoc V. Le},
    journal = {ArXiv},
    year    = {2022},
    volume  = {abs/2202.10447}
}
@software{peng_bo_2021_5196578,
    author    = {PENG Bo},
    title     = {BlinkDL/RWKV-LM: 0.01},
    month     = {aug},
    year      = {2021},
    publisher = {Zenodo},
    version   = {0.01},
    doi       = {10.5281/zenodo.5196578},
    url       = {https://doi.org/10.5281/zenodo.5196578}
}
Comments
  • einsum operation in Linear Attention Part

    einsum operation in Linear Attention Part

    Hi, Thanks a lot for your FLASH_pytorch, which helps a lot. I found that there are some differences from the paper in the Linear Attention Part: https://github.com/lucidrains/FLASH-pytorch/blob/main/flash_pytorch/flash_pytorch.py#L342-L343

    lin_kv = einsum('b g n d, b g n e -> b d e', lin_k, v) / n
lin_out = einsum('b g n d, b d e -> b g n e', lin_q, lin_kv)

    the lin_kv is three-dim (bde) And the code in the paper is

    lin_kv = tf.einsum('bhke,bgh→bgke', lin_kv, mask) 
linear = tf.einsum('bgnk,bgke→bgne', lin_q, lin_kv)

    the lin_kv is four-dim (bgke) It seems that the two ways are not equivalent.

    Looking forward to your reply. Best,

    opened by ShomyLiu 5
  • mask error

    mask error 

    x = torch.randint(0, 20000, (1, 1024))
mask = x.ne(0)
logits = model(x, mask=mask)

    RuntimeError: The size of tensor a (1024) must match the size of tensor b (128) at non-singleton dimension 2

    opened by keyunluo 1
  • Speed on TPU

    Speed on TPU

    Hi, Thanks for the code! I test it on Google TPU v3, the training speed seems slower than my expectation. Maybe there is some operation which is not lower on TPU.

    opened by magicknight 0
  • About the

    About the "shift_tokens"

    Thank you for your amazing code.

    In the class of FLASH, I find a flag: shift_tokens, and the corresponding code is as following: if self.shift_tokens: x_shift, x_pass = normed_x.chunk(2, dim = -1) x_shift = F.pad(x_shift, (0, 0, 1, -1), value = 0.) normed_x = torch.cat((x_shift, x_pass), dim = -1)

    Assume we have normed_x in the shape [1024, 512], the x_shift/x_pass is the shape of [1024, 256]. Then it adds a row (with all 0 value) and remove the last row in the x_shift, and concat x_shift and x_pass to get the normed_x.

    In my opinion, the F.pad operation will make the row in x_shift and x_pass do not match again.

    May I know why it works?

    Kang

    opened by kangzhao2 1
  • Cross-Attention?

    Cross-Attention?

    Hi, @lucidrains. Thank you for sharing this excellent implementation with us all! Do you have any thoughts as to what changes would need to be made to make cross-attention possible with your FLASH model?

    opened by amorehead 2
Owner
Phil Wang
Working with Attention. It's all we need
Phil Wang
GitHub Repository
Official implementation for the paper: Multi-label Classification with Partial Annotations using Class-aware Selective Loss

Multi-label Classification with Partial Annotations using Class-aware Selective Loss Paper | Pretrained models Official PyTorch Implementation Emanuel

99 Dec 27, 2022
Easy genetic ancestry predictions in Python

ezancestry Easily visualize your direct-to-consumer genetics next to 2500+ samples from the 1000 genomes project. Evaluate the performance of a custom

Kevin Arvai 38 Jan 02, 2023
A flexible ML framework built to simplify medical image reconstruction and analysis experimentation.

meddlr Getting Started Meddlr is a config-driven ML framework built to simplify medical image reconstruction and analysis problems. Installation To av

Arjun Desai 36 Dec 16, 2022
Code for paper "Which Training Methods for GANs do actually Converge? (ICML 2018)"

GAN stability This repository contains the experiments in the supplementary material for the paper Which Training Methods for GANs do actually Converg

Lars Mescheder 885 Jan 01, 2023
[NeurIPS 2020] Official repository for the project "Listening to Sound of Silence for Speech Denoising"

Listening to Sounds of Silence for Speech Denoising Introduction This is the repository of the "Listening to Sounds of Silence for Speech Denoising" p

Henry Xu 40 Dec 20, 2022
AEI: Actors-Environment Interaction with Adaptive Attention for Temporal Action Proposals Generation

AEI: Actors-Environment Interaction with Adaptive Attention for Temporal Action Proposals Generation A pytorch-version implementation codes of paper:

11 Dec 13, 2022
Structure Information is the Key: Self-Attention RoI Feature Extractor in 3D Object Detection

Structure Information is the Key: Self-Attention RoI Feature Extractor in 3D Object Detection abstract:Unlike 2D object detection where all RoI featur

DK. Zhang 2 Oct 07, 2022
Python library for computer vision labeling tasks. The core functionality is to translate bounding box annotations between different formats-for example, from coco to yolo.

PyLabel pip install pylabel PyLabel is a Python package to help you prepare image datasets for computer vision models including PyTorch and YOLOv5. I

PyLabel Project 176 Jan 01, 2023
Action Recognition for Self-Driving Cars

Action Recognition for Self-Driving Cars This repo contains the codes for the 2021 Fall semester project "Action Recognition for Self-Driving Cars" at

VITA lab at EPFL 3 Apr 07, 2022
This repo contains the official code of our work SAM-SLR which won the CVPR 2021 Challenge on Large Scale Signer Independent Isolated Sign Language Recognition.

Skeleton Aware Multi-modal Sign Language Recognition By Songyao Jiang, Bin Sun, Lichen Wang, Yue Bai, Kunpeng Li and Yun Fu. Smile Lab @ Northeastern

Isen (Songyao Jiang) 128 Dec 08, 2022
An index of algorithms for learning causality with data

awesome-causality-algorithms An index of algorithms for learning causality with data. Please cite our survey paper if this index is helpful. @article{

Ruocheng Guo 2.3k Jan 08, 2023
Code for the upcoming CVPR 2021 paper

The Temporal Opportunist: Self-Supervised Multi-Frame Monocular Depth Jamie Watson, Oisin Mac Aodha, Victor Prisacariu, Gabriel J. Brostow and Michael

Niantic Labs 496 Dec 30, 2022
A library of scripts that interact with the PythonTurtle module to create games, drawings, and more

TurtleLib TurtleLib is a library of scripts that interact with the PythonTurtle module to create games, drawings, and more! Using the Scripts Copy or

1 Jan 15, 2022
Simple-Neural-Network From Scratch in Python

Simple-Neural-Network From Scratch in Python This is a simple Neural Network created without any Machine Learning Libraries. The only dependencies are

Aum Shah 1 Dec 28, 2021
Revisiting Discriminator in GAN Compression: A Generator-discriminator Cooperative Compression Scheme (NeurIPS2021)

Revisiting Discriminator in GAN Compression: A Generator-discriminator Cooperative Compression Scheme (NeurIPS2021) Overview Prerequisites Linux Pytho

Shaojie Li 34 Mar 31, 2022
Voice Gender Recognition

In this project it was used some different Machine Learning models to identify the gender of a voice (Female or Male) based on some specific speech and voice attributes.

Anne Livia 1 Jan 27, 2022
Build and run Docker containers leveraging NVIDIA GPUs

NVIDIA Container Toolkit Introduction The NVIDIA Container Toolkit allows users to build and run GPU accelerated Docker containers. The toolkit includ

NVIDIA Corporation 15.6k Jan 01, 2023
Library for implementing reservoir computing models (echo state networks) for multivariate time series classification and clustering.

Framework overview This library allows to quickly implement different architectures based on Reservoir Computing (the family of approaches popularized

Filippo Bianchi 249 Dec 21, 2022
PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.

PySlowFast PySlowFast is an open source video understanding codebase from FAIR that provides state-of-the-art video classification models with efficie

Meta Research 5.3k Jan 03, 2023
2.86% and 15.85% on CIFAR-10 and CIFAR-100

Shake-Shake regularization This repository contains the code for the paper Shake-Shake regularization. This arxiv paper is an extension of Shake-Shake

Xavier Gastaldi 294 Nov 22, 2022