[{"data":1,"prerenderedAt":59},["ShallowReactive",2],{"maker-xaskasdf":3},[4,35],{"id":5,"name":6,"line":7,"ai":8,"cat":9,"year":10,"points":11,"comments":12,"rec":13,"s":14,"top":15,"kw":16,"maker":22,"domain":23,"tier":24,"attempts":25,"strip":26,"url":27,"mix":28,"praise":32,"objection":33,"sub":34},47104667,"ntransformer","Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPU",false,"ai",2026,395,101,"curious",-0.01,3,[17,18,19,20,21],"llm inference","nvme offload","gpudirect storage","consumer gpu","llama 70b","xaskasdf","github.com","legend",0,"3303130001331533330303333330313531003350","https:\u002F\u002Fgithub.com\u002Fxaskasdf\u002Fntransformer",[29,30,25,31],7,28,10,"Impressive systems engineering with tiered caching","Around 0.2 to 0.5 tokens per second is too slow for interactive use","local llm inference",{"id":36,"name":37,"line":38,"ai":8,"cat":9,"year":10,"points":39,"comments":40,"rec":41,"s":42,"top":43,"kw":44,"maker":22,"domain":23,"tier":50,"attempts":25,"strip":51,"url":52,"mix":53,"praise":56,"objection":57,"sub":58},47470405,"ps2-llm","I ran a language model on a PS2",46,12,"loved",0.27,30,[45,46,47,48,49],"ps2 llm","language model","quantized inference","emotion engine","cd-rom streaming","shown","33350030335055","https:\u002F\u002Fgithub.com\u002Fxaskasdf\u002Fps2-llm",[54,55,25,25],4,6,"Streaming weights from CD-ROM is a clever constraint hack","Inference speed is unclear; tok\u002Fsec counter reported zero","language model inference",1794036822364]