Real-time LLM Inference on Standard GPUs: 3k tokens/s per request

111 points | by NicoConstant 6 hours ago

58 comments