block recurrent transformer pytorch下載 - block recurrent transformer pytorch原始碼下載

block recurrent transformer pytorch

Ai源碼

0.4.4

下載

塊循環變壓器 - Pytorch

塊循環變壓器的實作 - Pytorch。這篇論文的亮點在於它能夠記住多達 60k 個標記之前的內容。

該設計是針對循環變壓器研究線的 SOTA。

它還將包括 Flash Attention 以及使用本文中的想法的多達 250k 令牌的路由記憶

欣賞

Stability.ai 慷慨贊助工作與開源尖端人工智慧研究

安裝

$ pip install block-recurrent-transformer-pytorch

用法

 import torch
from block_recurrent_transformer_pytorch import BlockRecurrentTransformer

model = BlockRecurrentTransformer (
    num_tokens = 20000 ,             # vocab size
    dim = 512 ,                      # model dimensions
    depth = 6 ,                      # depth
    dim_head = 64 ,                  # attention head dimensions
    heads = 8 ,                      # number of attention heads
    max_seq_len = 1024 ,             # the total receptive field of the transformer, in the paper this was 2 * block size
    block_width = 512 ,              # block size - total receptive field is max_seq_len, 2 * block size in paper. the block furthest forwards becomes the new cached xl memories, which is a block size of 1 (please open an issue if i am wrong)
    num_state_vectors = 512 ,        # number of state vectors, i believe this was a single block size in the paper, but can be any amount
    recurrent_layers = ( 4 ,),        # where to place the recurrent layer(s) for states with fixed simple gating
    use_compressed_mem = False ,     # whether to use compressed memories of a single block width, from https://arxiv.org/abs/1911.05507
    compressed_mem_factor = 4 ,      # compression factor of compressed memories
    use_flash_attn = True           # use flash attention, if on pytorch 2.0
)

seq = torch . randint ( 0 , 2000 , ( 1 , 1024 ))

out , mems1 , states1 = model ( seq )
out , mems2 , states2 = model ( seq , xl_memories = mems1 , states = states1 )
out , mems3 , states3 = model ( seq , xl_memories = mems2 , states = states2 )

在Enwik8上測試

首先pip install -r requirements.txt ，然後

$ python train.py

托多

引文

 @article { Hutchins2022BlockRecurrentT ,
    title   = { Block-Recurrent Transformers } ,
    author  = { DeLesley S. Hutchins and Imanol Schlag and Yuhuai Wu and Ethan Dyer and Behnam Neyshabur } ,
    journal = { ArXiv } ,
    year    = { 2022 } ,
    volume  = { abs/2203.07852 }
}

 @article { Shazeer2019FastTD ,
    title   = { Fast Transformer Decoding: One Write-Head is All You Need } ,
    author  = { Noam M. Shazeer } ,
    journal = { ArXiv } ,
    year    = { 2019 } ,
    volume  = { abs/1911.02150 }
}

 @inproceedings { Sun2022ALT ,
    title     = { A Length-Extrapolatable Transformer } ,
    author    = { Yutao Sun and Li Dong and Barun Patra and Shuming Ma and Shaohan Huang and Alon Benhaim and Vishrav Chaudhary and Xia Song and Furu Wei } ,
    year      = { 2022 }
}

 @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 }
}

 @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 }
}