Mamba2
Benchmarks
TaskDatasetSOTA resultTrendResults
Mamba2 8B
16Compression Time
3
Mamba2-8B on Nano 8G (inference)
55.3TPOT (Time Per Output Token)
2
Mamba2-8B Jetson Orin Nano 8G (512 prefill 512 generation tokens) (inference)
153.64Energy (J/req)
2