Hacker News new | ask | show | jobs
by rohansood15 92 days ago
The paper is about vector quantization, which affects KV cache not model weights/sizes.