Unduh simple hierarchical transformer - pengunduhan kode sumber simple hierarchical transformer

Transformator Hirarki Sederhana

Eksperimen seputar ide sederhana untuk mendorong beberapa model pengkodean prediktif hierarkis dalam GPT. Ini sangat sederhana, mungkin tidak berhasil. Namun sekali lagi, kemajuan pembelajaran mendalam dibangun di atas fondasi ide-ide sederhana. Layak untuk dicoba.

Sejauh ini, ide tersebut telah lolos uji lakmus dari seorang teman peneliti. Akan menyelesaikannya sekitar minggu depan. Jika tidak berhasil, saya akan meninggalkan hasil eksperimen negatif serta penyimpanannya, dan mungkin beberapa mahasiswa PhD dapat mengembangkannya.

Pembaruan: Saya pikir ini berfungsi?

Apresiasi

StabilityAI atas sponsor untuk melaksanakan penelitian independen ini
? Huggingface untuk perpustakaan akselerasi mereka

Memasang

 $ pip install simple-hierarchical-transformer

Penggunaan

Tiga hierarki, semuanya melayani prediksi token berikutnya

 import torch
from simple_hierarchical_transformer import HierarchicalTransformer

model = HierarchicalTransformer (
    num_tokens = 20000 ,                # number of tokens
    dim = 512 ,                         # model dimensions
    depth = 6 ,                         # depth
    dim_head = 64 ,                     # dimension per attention head
    heads = 8 ,                         # attention heads
    seq_len = 2048 ,                    # sequence lengths
    hierarchies = ( 1 , 2 , 8 ),           # hierarchies - here we have 1x (like in a regular transformer), then 2x and 8x compressed hierarchical tokens that undergo their own transformer blocks. information is pooled into one hierarchy at each layer
    window_sizes = ( 32 , 64 , None )      # local attention window sizes - the idea is that the higher hierarchies can pass distant information to the local one. None stands for full receptive field. Setting 0 would turn off attention at this hierarchy altogether (while token shift will still be in effect in each layer)
)

ids = torch . randint ( 0 , 20000 , ( 1 , 2048 ))

loss , _ = model ( ids , return_loss = True )
loss . backward ()

# after much training

logits = model ( ids )

Dengan tidak menentukan hierarchies dan window_sizes , pada dasarnya Anda menggunakan transformator autoregresif biasa dengan perhatian pada seluruh panjang urutan.

 # non-hierarchical transformer

model = HierarchicalTransformer (
    num_tokens = 20000 ,
    dim = 512 ,
    depth = 8 ,
    dim_head = 64 ,
    heads = 8 ,
    seq_len = 2048 ,
    hierarchies = 1 ,        # implied 1 if not set
    window_sizes = None     # implied None (full sequence length) if not set
)

Sekarang sesuatu yang lebih kompleks. Eksperimen menunjukkan bahwa saat Anda mengompresi hierarki, Anda memerlukan dimensi model yang lebih besar untuk kapasitas yang sesuai.

 model = HierarchicalTransformer (
    num_tokens = 256 ,
    dim = ( 128 , 256 , 512 , 1024 ),
    depth = 8 ,
    seq_len = 1024 ,
    use_flash_attn = True ,
    ff_mult = ( 2 , 2 , 4 , 4 ),
    dim_head = ( 16 , 32 , 64 , 64 ),
    heads = ( 2 , 4 , 8 , 8 ),
    hierarchies = ( 1 , 2 , 4 , 16 ),
    hierarchical_stride = ( 1 , 1 , 1 , 8 ),  # this would determine the stride when compressing, and when concatting the hierarchical tokens to the fine tokens, the past tokens will be repeated this amount of time. causality is not violated as using the trick from hourglass transformers where sequence is shifted by compression factor - 1. recommend sticking with 1 except for highly compressed hierarchies, as it becomes very uncompetitive with baseline and generations look off
    window_sizes = ( 16 , 32 , 64 , None )
). cuda ()

# hierarchies
# 1x - dim 128 - attention (2 heads, 16 dim, receptive field 16)
# 2x - dim 256 - attention (4 heads, 32 dim, receptive field 32)
# 4x - dim 512 - attention (8 heads, 64 dim, receptive field 64)
# 8x - dim 1024 - attention (8 heads, 64 dim, receptive field of all)

Semua yang harus dilakukan

Kutipan

Ide yang paling mendekati adalah trafo jam pasir.

Dan minat saya yang baru terhadap pendekatan hierarki muncul dari membaca ini.

 @article { Nawrot2021HierarchicalTA ,
    title   = { Hierarchical Transformers Are More Efficient Language Models } ,
    author  = { Piotr Nawrot and Szymon Tworkowski and Michal Tyrolski and Lukasz Kaiser and Yuhuai Wu and Christian Szegedy and Henryk Michalewski } ,
    journal = { ArXiv } ,
    year    = { 2021 } ,
    volume  = { abs/2110.13711 }
}

 @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 }
}

 @misc { su2021roformer ,
    title   = { RoFormer: Enhanced Transformer with Rotary Position Embedding } ,
    author  = { Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu } ,
    year    = { 2021 } ,
    eprint  = { 2104.09864 } ,
    archivePrefix = { arXiv } ,
    primaryClass = { cs.CL }
}

 @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 }
}

 @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 }
}

 @article { Piergiovanni2023Mirasol3BAM ,
    title   = { Mirasol3B: A Multimodal Autoregressive model for time-aligned and contextual modalities } ,
    author  = { A. J. Piergiovanni and Isaac Noble and Dahun Kim and Michael S. Ryoo and Victor Gomes and Anelia Angelova } ,
    journal = { ArXiv } ,
    year    = { 2023 } ,
    volume  = { abs/2311.05698 } ,
    url     = { https://api.semanticscholar.org/CorpusID:265129010 }
}