XAMBA makes Mamba and Mamba-2 state-space models run up to 4.8x faster on an Intel AI PC NPU by replacing CumSum and ReduceSum with matrix operations and activations with piecewise linear approximations.
Intel® Core™ Ultra series 1 product brief , 2024 b
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
XAMBA: Enabling Efficient State Space Models on Resource-Constrained Neural Processing Units
XAMBA makes Mamba and Mamba-2 state-space models run up to 4.8x faster on an Intel AI PC NPU by replacing CumSum and ReduceSum with matrix operations and activations with piecewise linear approximations.