Article URL: https://grapheneos.social/@GrapheneOS/114327666433966529
Comments URL: https://news.ycombinator.com/item?id=43669185
Points: 40
# Comments: 3
Erstellt
1mo
|
13.04.2025, 02:30:10
Melden Sie sich an, um einen Kommentar hinzuzufügen
Andere Beiträge in dieser Gruppe

I discovered that in LLM inference, keys and values in the KV cache have very different quantization sensitivities. Keys need higher precision than values to maintain quality.
I patched llama.cp
Article URL: https://clojurescript.org/news/2025-05-16-release




Article URL: https://bobacollection.staxmuseum.org/
Comments URL: https://news.y