FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution Paper • 2608.16157 • Published Aug 17 • 110
M2RNN Collection Note that the 7B models are MoE with 1.1B active parameters and 400M models are dense models • 14 items • Updated Mar 12 • 7
M2RNN Collection Note that the 7B models are MoE with 1.1B active parameters and 400M models are dense models • 14 items • Updated Mar 12 • 7