ollama v0.34.2-rc2: mlxrunner: Release freed KV buffers during speculative decode

The decode loop releases MLX's pool of freed buffers every 256 generated tokens, which is also how often the KV cache grows and drops its previous,…

The decode loop releases MLX's pool of freed buffers every 256 generated tokens, which is also how often the KV cache grows and drops its previous, smaller buffers. The check fires only when the token count lands exactly on a multiple of 256. Speculative decoding emits several…

Read the original source — github.com

release · Shared by tscosj

0 comments

No comments yet.