server: use pos_next instead of n_tokens for m-rope#22439

ggerganov merged 1 commit into

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

am17an:server-spec-pos-rollbackam17an/llama.cpp:server-spec-pos-rollbackCopy head branch name to clipboard