youtube.com/watch?v=0fWF...
#1bit #LLM #noGPU #noNPU #justCPU
The prospect of functional #edgeLLM in a #RISC-V instruction extension (or M chip) already may have popped #AIbubble. The #TurboQuant #KVC optimization was less disruptive than this approach to small AIs useful for #privacy or #latency.
2
0
1
0