Paper published on a website (Scientific congresses and symposiums)
Parallelizing Autoregressive Generation with Variational State Space Models
Lambrechts, Gaspard; Claes, Yann; Geurts, Pierre et al.
2024ICML Workshop on Next Generation of Sequence Modeling Architectures
Peer reviewed
 

Files


Full Text
vssm-ngsm.pdf
Author postprint (1.37 MB)
Download
Annexes
vssm-poster.pdf
(416.64 kB) Creative Commons License - Attribution, ShareAlike
Poster
Download

All documents in ORBi are protected by a user license.

Send to



Details



Keywords :
Parallel; Autoregressive; Generation; VAE; SSM; VSSM
Abstract :
[en] Attention-based models such as Transformers and recurrent models like state space models (SSMs) have emerged as successful methods for autoregressive sequence modeling. Although both enable parallel training, none enable parallel generation due to their autoregressiveness. We propose the variational SSM (VSSM), a variational autoencoder (VAE) where both the encoder and decoder are SSMs. Since sampling the latent variables and decoding them with the SSM can be parallelized, both training and generation can be conducted in parallel. Moreover, the decoder recurrence allows generation to be resumed without reprocessing the whole sequence. Finally, we propose the autoregressive VSSM that can be conditioned on a partial realization of the sequence, as is common in language generation tasks. Interestingly, the autoregressive VSSM still enables parallel generation. We highlight on toy problems (MNIST, CIFAR) the empirical gains in speed-up and show that it competes with traditional models in terms of generation quality (Transformer, Mamba SSM).
Disciplines :
Computer science
Author, co-author :
Lambrechts, Gaspard  ;  Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
Claes, Yann   ;  Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
Geurts, Pierre  ;  Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
Ernst, Damien  ;  Université de Liège - ULiège > Montefiore Institute of Electrical Engineering and Computer Science
 These authors have contributed equally to this work.
Language :
English
Title :
Parallelizing Autoregressive Generation with Variational State Space Models
Publication date :
July 2024
Event name :
ICML Workshop on Next Generation of Sequence Modeling Architectures
Event place :
Vienne, Austria
Event date :
26/07/2024
Audience :
International
Peer reviewed :
Peer reviewed
Tags :
CÉCI : Consortium des Équipements de Calcul Intensif
Tier-1 supercomputer
Available on ORBi :
since 12 July 2024

Statistics


Number of views
84 (16 by ULiège)
Number of downloads
23 (5 by ULiège)

Bibliography


Similar publications



Sorry the service is unavailable at the moment. Please try again later.
Contact ORBi