Download CoLT5 attention - Download do código-fonte CoLT5 attention

Atenção CoLT5 - Pytorch

Implementação da atenção eficiente roteada condicionalmente na arquitetura CoLT5 proposta, em Pytorch.

Eles usaram a descida de coordenadas deste artigo (algoritmo principal originalmente de Wright et al) para rotear um subconjunto de tokens para ramos 'mais pesados' dos blocos de feedforward e de atenção.

Atualização: não tenho certeza de como as pontuações normalizadas de roteamento para os valores-chave são usadas. Fiz alguns improvisos aí, escalando os valores projetados, mas se você acha que sabe a resposta, abra um fascículo

Atualização 2: parece funcionar bem com a improvisação acima

Apreciação

Stability.ai pelo generoso patrocínio para trabalhar em pesquisas de ponta em inteligência artificial
einops por facilitar minha vida
Triton por me permitir acelerar a descida coordenada com uma implementação fundida em apenas 2 dias, poupando-me de ter que escrever mil linhas de código CUDA

Instalar

$ pip install colt5-attention

Uso

 import torch

from colt5_attention import (
    ConditionalRoutedFeedForward ,
    ConditionalRoutedAttention ,
    ConditionalRoutedTransformerBlock
)

# mock input, say it is 32768 length

tokens = torch . randn ( 2 , 32768 , 512 )
mask = torch . ones ( 2 , 32768 ). bool ()  # can handle variable lengthed sequences

# feedforward

ff = ConditionalRoutedFeedForward (
    dim = 512 ,
    light_ff_mult = 0.5 ,      # hidden dimension ratio of light branch
    heavy_ff_mult = 4 ,        # hidden dimension ratio of heavy branch
    num_heavy_tokens = 1024   # heavy branch receives only 1024 routed tokens of 32768
)

ff_out = ff ( tokens , mask = mask )  # (2, 32768, 512) - light and heavy branch summed

# attention

attn = ConditionalRoutedAttention (
    dim = 512 ,
    light_dim_head = 64 ,       # attention head dimension of light branch
    light_heads = 8 ,           # number of attention heads for light branch
    light_window_size = 128 ,   # local attention receptive field for light
    heavy_dim_head = 64 ,       # attention head dimension of heavy branch
    heavy_heads = 8 ,           # number of attention heads for heavy branch
    num_heavy_tokens_q = 1024 , # heavy branch receives only 1024 routed tokens of 32768
    num_heavy_tokens_kv = 1024 # heavy branch receives only 1024 routed tokens of 32768
)

attn_out = attn ( tokens , mask = mask ) # (2, 32768, 512) - light and heavy branch summed

# both attention and feedforward with residual
# the complete transformer block
# a stack of these would constitute the encoder of CoLT5

block = ConditionalRoutedTransformerBlock (
    dim = 512 ,
    light_dim_head = 64 ,
    light_heads = 8 ,
    light_window_size = 128 ,
    heavy_dim_head = 64 ,
    heavy_heads = 8 ,
    light_ff_mult = 0.5 ,
    heavy_ff_mult = 4 ,
    num_heavy_ff_tokens = 1024 ,
    num_heavy_attn_tokens_q = 1024 ,
    num_heavy_attn_tokens_kv = 1024
)

block_out = block ( tokens , mask = mask ) # (2, 32768, 512)

Também foi incluída uma variação da atenção condicionalmente roteada para atenção cruzada, para ser testada com memórias de contexto longas em um transformer-xl

 import torch
from colt5_attention import ConditionalRoutedCrossAttention

# mock input, let us say it is a transformer of 1024 length attending to 1 million context past memories

tokens = torch . randn ( 1 , 1024 , 512 ). cuda ()
tokens_mask = torch . ones ( 1 , 1024 ). bool (). cuda ()

memories = torch . randn ( 1 , 1_048_576 , 512 ). cuda ()
memories_mask = torch . ones ( 1 , 1_048_576 ). bool (). cuda ()

# conditionally routed cross attention

cross_attn = ConditionalRoutedCrossAttention (
    dim = 512 ,
    dim_head = 64 ,
    heads = 8 ,
    num_tokens_q = 512 ,         # only 512 routed from 1024
    num_tokens_kv = 1024 ,       # only 1024 routed from 1 million
    kv_routing_tokens = 2 ,      # say you want 2 routing tokens to route different sets of key / values to the queries. 4 attention heads will be allocated to each routed set in this example (8 / 2)
    use_triton = True ,          # use cuda kernel
    route_block_size = 131072   # route in blocks of 131072
). cuda ()

cross_attn_out = cross_attn (
    tokens ,
    context = memories ,
    mask = tokens_mask ,
    context_mask = memories_mask
)

cross_attn_out . shape # (1, 1024, 512) - same as tokens

Este repositório também possui uma versão improvisada para atenção autorregressiva. A forma como isso foi conseguido foi visualizando a sequência em janelas. Cada janela só pode atender a janelas de chaves/valores do passado. A atenção local do ramo de luz cobre a atenção intrajanela.

A descida coordenada é viabilizada através de um kernel CUDA escrito em Triton. Finalmente, para que a geração autoregressiva funcionasse bem, tive que garantir que os tokens não roteados (para consultas) produzissem uma incorporação de saída aprendida em vez de apenas zeros.

Atualmente estou vendo diferenças ocasionais entre os gradientes (tão altas quanto 1e-1 para uma fração muito pequena de elementos) quando o número de iterações excede 20. No entanto, o enwik8 parece treinar bem e posso ver os efeitos do roteamento. O treinamento também é surpreendentemente estável

ex.

 import torch
from colt5_attention import ConditionalRoutedAutoregressiveAttention

# mock input, say it is 8192 length

tokens = torch . randn ( 2 , 8192 , 512 ). cuda ()

# attention

attn = ConditionalRoutedAutoregressiveAttention (
    dim = 512 ,
    light_dim_head = 64 ,          # attention head dimension of light branch
    light_heads = 8 ,              # number of attention heads for light branch
    light_window_size = 128 ,      # local attention receptive field for light
    heavy_window_size = 128 ,      # the windowing for the routed heavy attention, by default, will be equal to the light window size. be aware if this is any greater than the light window size, there may be tokens that would be missed by attention
    heavy_dim_head = 64 ,          # attention head dimension of heavy branch
    heavy_heads = 8 ,              # number of attention heads for heavy branch
    num_heavy_tokens_q = 32 ,      # heavy branch receives only 32 out of 128 of the windowed queries (1024 query tokens total)
    num_heavy_tokens_kv = 1024 ,   # heavy branch receives only 1024 routed tokens for key-values
    num_routed_kv = 2 ,            # one can split the attention heads so that groups of heads attend to different sets of key - values (2 routing tokens in this case)
    use_triton = True ,            # will need to use Triton for this to be viable, otherwise it is too slow and memory efficient with the number of iterations
    use_flash_attn = True         # use flash attention in heavy branch
). cuda ()

attn_out = attn ( tokens ) + tokens # (2, 8192, 512) - output of attention with residual (prenorm is included)

Finalmente, este repositório contém uma versão para mapas de recursos de imagem. Normalmente, muitos artigos de pesquisa não conseguem prestar atenção em mapas de características de imagem com dimensões maiores que 32 por 32. Essa atenção direcionada usará um patch de janela local para o ramo leve e atenção direcionada para o ramo pesado.

ex.

 import torch
from colt5_attention import ConditionalRoutedImageAttention

attn = ConditionalRoutedImageAttention (
    dim = 32 ,
    light_dim_head = 64 ,       # attention head dimension of light branch
    light_heads = 8 ,           # number of attention heads for light branch
    light_window_size = 32 ,    # height and width of local window attention on the image feature map
    channel_first = True ,      # whether to accept images with channel first than last
    heavy_dim_head = 64 ,       # attention head dimension of heavy branch
    heavy_heads = 8 ,           # number of attention heads for heavy branch
    num_heavy_tokens_q = 1024 , # heavy branch receives only 1024 routed tokens of 65536
    num_heavy_tokens_kv = 1024 # heavy branch receives only 1024 routed tokens of 65536
). cuda ()

fmap = torch . randn ( 1 , 32 , 256 , 256 ). cuda () # image feature map is too large for attention, given 256 ^ 2  == 65536 tokens

out = attn ( fmap )

ViT simples usando descida coordenada roteada atenção e feedforward

 import torch
from colt5_attention . vit import ConditionalRoutedViT

vit = ConditionalRoutedViT (
    image_size = 256 ,                # image size
    patch_size = 32 ,                 # patch size
    num_classes = 1000 ,              # number of output classes
    dim = 1024 ,                      # feature dimension
    depth = 6 ,                       # depth
    attn_num_heavy_tokens_q = 16 ,    # number of routed queries for heavy attention
    attn_num_heavy_tokens_kv = 16 ,   # number of routed key/values for heavy attention
    attn_heavy_dim_head = 64 ,        # dimension per attention head for heavy
    attn_heavy_heads = 8 ,            # number of attention heads for heavy
    attn_light_window_size = 4 ,      # the local windowed attention for light branch
    attn_light_dim_head = 32 ,        # dimension per head for local light attention
    attn_light_heads = 4 ,            # number of attention heads for local windowed attention
    ff_num_heavy_tokens = 16 ,        # number of tokens routed for heavy feedforward
    ff_heavy_mult = 4 ,               # the expansion factor of the heavy feedforward branch
    ff_light_mult = 2                # expansion factor of the light feedforward branch
)

images = torch . randn ( 1 , 3 , 256 , 256 )

logits = vit ( images ) # (1, 1000)

Topk diferenciável

Use um pequeno invólucro em torno da descida coordenada para topk diferenciável

 import torch
from colt5_attention import topk

x = torch . randn ( 1024 , 512 )

values , indices , coor_descent_values , gates = topk ( x , k = 10 , fused = True )

# you can either use the topk indices + gates, or use the values directly (values have already been multiplied with the gates within the function)

Pendência

Citações

 @inproceedings { Ainslie2023CoLT5FL ,
    title   = { CoLT5: Faster Long-Range Transformers with Conditional Computation } ,
    author  = { Joshua Ainslie and Tao Lei and Michiel de Jong and Santiago Ontan'on and Siddhartha Brahma and Yury Zemlyanskiy and David Uthus and Mandy Guo and James Lee-Thorp and Yi Tay and Yun-Hsuan Sung and Sumit Sanghai } ,
    year    = { 2023 }
}

 @article { Tillet2019TritonAI ,
    title   = { Triton: an intermediate language and compiler for tiled neural network computations } ,
    author  = { Philippe Tillet and H. Kung and D. Cox } ,
    journal = { Proceedings of the 3rd ACM SIGPLAN International Workshop on Machine Learning and Programming Languages } ,
    year    = { 2019 }
}

 @inproceedings { dao2022flashattention ,
    title     = { Flash{A}ttention: Fast and Memory-Efficient Exact Attention with {IO}-Awareness } ,
    author    = { Dao, Tri and Fu, Daniel Y. and Ermon, Stefano and Rudra, Atri and R{'e}, Christopher } ,
    booktitle = { Advances in Neural Information Processing Systems } ,
    year      = { 2022 }
}

 @article { Lei2023ConditionalAP ,
    title   = { Conditional Adapters: Parameter-efficient Transfer Learning with Fast Inference } ,
    author  = { Tao Lei and Junwen Bai and Siddhartha Brahma and Joshua Ainslie and Kenton Lee and Yanqi Zhou and Nan Du and Vincent Zhao and Yuexin Wu and Bo Li and Yu Zhang and Ming-Wei Chang } ,
    journal = { ArXiv } ,
    year    = { 2023 } ,
    volume  = { abs/2304.04947 }
}

 @article { Beyer2022BetterPV ,
    title   = { Better plain ViT baselines for ImageNet-1k } ,
    author  = { Lucas Beyer and Xiaohua Zhai and Alexander Kolesnikov } ,
    journal = { ArXiv } ,
    year    = { 2022 } ,
    volume  = { abs/2205.01580 }
}