An end-to-end Mamba edge accelerator using hardware-friendly approximations, INT8 quantization, and NAS achieves 4.95x-5.62x lower latency and 1.63x-19.9x smaller models than ViT/CNN baselines.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
eMamba: Efficient Acceleration Framework for Mamba Models in Edge Computing
An end-to-end Mamba edge accelerator using hardware-friendly approximations, INT8 quantization, and NAS achieves 4.95x-5.62x lower latency and 1.63x-19.9x smaller models than ViT/CNN baselines.