server: MTP layer kv-cache should respect draft type ctk#23646

am17an merged 1 commit into

ggml-org:masterggml-org/llama.cpp:masterfrom

am17an:mtp-kv-cache-fixam17an/llama.cpp:mtp-kv-cache-fixCopy head branch name to clipboard