recurrent memory transformer pytorch
0.7.0
Pytorch에서 순환 메모리 변환기(openreview) 구현. 그들은 최근 최소한 100만 개의 토큰에 걸쳐 정보를 복사할 수 있음을 보여주는 짧은 후속 보고서를 가지고 있었습니다.
RMT가 단지 Transformer-XL인 AdA보다 더 강력한 RL 에이전트를 만들 것이라는 점에는 의심의 여지가 없습니다. - 업데이트: Recurrent Action Transformer with Memory(RATE)
Yannic Kilcher 논문 리뷰
$ pip install recurrent-memory-transformer-pytorch import torch
from recurrent_memory_transformer_pytorch import RecurrentMemoryTransformer
model = RecurrentMemoryTransformer (
num_tokens = 20000 , # number of tokens
num_memory_tokens = 128 , # number of memory tokens, this will determine the bottleneck for information being passed to the future
dim = 512 , # model dimensions
depth = 6 , # transformer depth
causal = True , # autoregressive or not
dim_head = 64 , # dimension per head
heads = 8 , # heads
seq_len = 1024 , # sequence length of a segment
use_flash_attn = True # whether to use flash attention
)
x = torch . randint ( 0 , 256 , ( 1 , 1024 ))
logits1 , mem1 , _ = model ( x ) # (1, 1024, 20000), (1, 128, 512), None
logits2 , mem2 , _ = model ( x , mem1 ) # (1, 1024, 20000), (1, 128, 512), None
logits3 , mem3 , _ = model ( x , mem2 ) # (1, 1024, 20000), (1, 128, 512), None
# and so on ...XL 추억과 함께
import torch
from recurrent_memory_transformer_pytorch import RecurrentMemoryTransformer
model = RecurrentMemoryTransformer (
num_tokens = 20000 ,
num_memory_tokens = 128 ,
dim = 512 ,
depth = 6 ,
causal = True ,
dim_head = 64 ,
heads = 8 ,
seq_len = 1024 ,
use_flash_attn = True ,
use_xl_memories = True , # set this to True
xl_mem_len = 512 # can be shorter than the seq len - i think just having a bit of the past will prevent much of the RMT memories memorizing the immediate preceding text
)
x = torch . randint ( 0 , 256 , ( 1 , 1024 ))
logits1 , mem1 , xl_mem1 = model ( x ) # (1, 1024, 20000), (1, 128, 512), [(2, 1, 512, 512)]
logits2 , mem2 , xl_mem2 = model ( x , mem1 , xl_memories = xl_mem1 ) # (1, 1024, 20000), (1, 128, 512), [(2, 1, 512, 512)]
logits3 , mem3 , xl_mem3 = model ( x , mem2 , xl_memories = xl_mem2 ) # (1, 1024, 20000), (1, 128, 512), [(2, 1, 512, 512)]
# and so on ...터무니없이 긴 시퀀스로 훈련
import torch
from recurrent_memory_transformer_pytorch import (
RecurrentMemoryTransformer ,
RecurrentMemoryTransformerWrapper
)
model = RecurrentMemoryTransformer (
num_tokens = 256 ,
num_memory_tokens = 128 ,
dim = 512 ,
depth = 6 ,
seq_len = 1024 ,
use_flash_attn = True ,
causal = True
)
model = RecurrentMemoryTransformerWrapper ( model ). cuda ()
seq = torch . randint ( 0 , 256 , ( 4 , 65536 )). cuda () # absurdly long sequence, in reality, they curriculum learned this starting with 1 segment to about 7-8 segments
loss = model ( seq , memory_replay_backprop = True ) # memory efficient training from memformer paper 메모리 재생 역전파를 torch.function으로 옮기고 양방향으로 테스트한 다음 실제 문제에 대해 테스트합니다.
xl 메모리로 회전식 임베딩이 제대로 작동하도록 하세요.
xl 메모리 추가, 분리
회전식 임베딩, 절대 위치 임베딩을 끄고 토큰 이동을 추가하는 방법을 제공합니다.
추억을 인과적으로 가려지는 옵션을 선택하세요
memformer 논문의 메모리 재생 역전파 기술을 추가하세요.
상대 위치 인코딩
블록 순환 변압기
멤포머
@inproceedings { bulatov2022recurrent ,
title = { Recurrent Memory Transformer } ,
author = { Aydar Bulatov and Yuri Kuratov and Mikhail Burtsev } ,
booktitle = { Advances in Neural Information Processing Systems } ,
editor = { Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho } ,
year = { 2022 } ,
url = { https://openreview.net/forum?id=Uynr3iPhksa }
} @misc { bulatov2023scaling ,
title = { Scaling Transformer to 1M tokens and beyond with RMT } ,
author = { Aydar Bulatov and Yuri Kuratov and Mikhail S. Burtsev } ,
year = { 2023 } ,
eprint = { 2304.11062 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CL }
} @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 { shazeer2020glu ,
title = { GLU Variants Improve Transformer } ,
author = { Noam Shazeer } ,
year = { 2020 } ,
url = { https://arxiv.org/abs/2002.05202 }
} @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 { Wu2020MemformerAM ,
title = { Memformer: A Memory-Augmented Transformer for Sequence Modeling } ,
author = { Qingyang Wu and Zhenzhong Lan and Kun Qian and Jing Gu and Alborz Geramifard and Zhou Yu } ,
booktitle = { AACL/IJCNLP } ,
year = { 2020 }
} @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 }
} @misc { ding2021cogview ,
title = { CogView: Mastering Text-to-Image Generation via Transformers } ,
author = { Ming Ding and Zhuoyi Yang and Wenyi Hong and Wendi Zheng and Chang Zhou and Da Yin and Junyang Lin and Xu Zou and Zhou Shao and Hongxia Yang and Jie Tang } ,
year = { 2021 } ,
eprint = { 2105.13290 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CV }
} @software { Dayma_DALLE_Mini_2021 ,
author = { Dayma, Boris and Patil, Suraj and Cuenca, Pedro and Saifullah, Khalid and Abraham, Tanishq and Lê Khắc, Phúc and Melas, Luke and Ghosh, Ritobrata } ,
doi = { 10.5281/zenodo.5146400 } ,
license = { Apache-2.0 } ,
month = { jul } ,
title = { {DALL·E Mini} } ,
url = { https://github.com/borisdayma/dalle-mini } ,
version = { v0.1-alpha } ,
year = { 2021 } } @inproceedings { anonymous2022normformer ,
title = { NormFormer: Improved Transformer Pretraining with Extra Normalization } ,
author = { Anonymous } ,
booktitle = { Submitted to The Tenth International Conference on Learning Representations } ,
year = { 2022 } ,
url = { https://openreview.net/forum?id=GMYWzWztDx5 } ,
note = { under review }
} @misc { ding2021erniedoc ,
title = { ERNIE-Doc: A Retrospective Long-Document Modeling Transformer } ,
author = { Siyu Ding and Junyuan Shang and Shuohuan Wang and Yu Sun and Hao Tian and Hua Wu and Haifeng Wang } ,
year = { 2021 } ,
eprint = { 2012.15688 } ,
archivePrefix = { arXiv } ,
primaryClass = { cs.CL }
} @article { Zhu2024HyperConnections ,
title = { Hyper-Connections } ,
author = { Defa Zhu and Hongzhi Huang and Zihao Huang and Yutao Zeng and Yunyao Mao and Banggu Wu and Qiyang Min and Xun Zhou } ,
journal = { ArXiv } ,
year = { 2024 } ,
volume = { abs/2409.19606 } ,
url = { https://api.semanticscholar.org/CorpusID:272987528 }
} @inproceedings { Zhou2024ValueRL ,
title = { Value Residual Learning For Alleviating Attention Concentration In Transformers } ,
author = { Zhanchao Zhou and Tianyi Wu and Zhiyun Jiang and Zhenzhong Lan } ,
year = { 2024 } ,
url = { https://api.semanticscholar.org/CorpusID:273532030 }
}
