3 ms·i run llama2 13B models with 4-6 k-quantized oin a 3060 with 12Gb VRamby SubiculumCode 3y agoi run llama2 13B models with 4-6 k-quantized oin a 3060 with 12Gb VRam