simple hierarchical transformer herunterladen - simple hierarchical transformer Quellcode herunterladen

Einfacher hierarchischer Transformator

Experimente rund um eine einfache Idee zur Einführung mehrerer hierarchischer prädiktiver Codierungsmodelle innerhalb eines GPT. Es ist so einfach, dass es möglicherweise nicht funktioniert. Andererseits basieren tiefgreifende Lernfortschritte auf der Grundlage einfacher Ideen. Einen Versuch wert.

Bisher hat die Idee den Lackmustest eines Forschungsfreundes bestanden. Werde es in der nächsten Woche oder so fertigstellen. Wenn es nicht klappt, lasse ich die negativen experimentellen Ergebnisse sowie das Repository liegen, und vielleicht kann ein Doktorand darauf aufbauen.

Update: Ich denke, es funktioniert?

Anerkennung

StabilityAI für die Unterstützung bei der Durchführung dieser unabhängigen Forschung
? Huggingface für ihre Beschleunigungsbibliothek

Installieren

 $ pip install simple-hierarchical-transformer

Verwendung

Drei Hierarchien, die alle das nächste Token vorhersagen

 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 )

Wenn Sie hierarchies und window_sizes nicht angeben, verwenden Sie grundsätzlich standardmäßig einen regulären autoregressiven Transformator mit Berücksichtigung der gesamten Sequenzlänge.

 # 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
)

Nun etwas Komplexeres. Experimente zeigen, dass Sie mit zunehmender Komprimierung der Hierarchien größere Modelldimensionen für eine entsprechende Kapazität benötigen.

 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)

Todo

Zitate

Die naheliegendste Idee wären Sanduhrtransformatoren.

Und mein erneutes Interesse an hierarchischen Ansätzen entstand durch die Lektüre dieser Zeilen.

 @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 }
}