Attamba compresses chunks of tokens with state-space models and runs attention over the compressed key-value states, improving WikiText2 perplexity by up to 24% over a reduced-footprint transformer baseline at small scale.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Attamba: Attending To Multi-Token States
Attamba compresses chunks of tokens with state-space models and runs attention over the compressed key-value states, improving WikiText2 perplexity by up to 24% over a reduced-footprint transformer baseline at small scale.