From 8c391fe120cbb4161d2cfe3c32eb07b0ba811782 Mon Sep 17 00:00:00 2001 From: Frank Date: Tue, 18 Aug 2026 16:11:58 -0700 Subject: [PATCH] Reduce ctxcp from 32 to 16 to free ~16 GiB VRAM headroom and prevent OOM crashes during prompt processing --- scripts/serve-frank-large.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/scripts/serve-frank-large.sh b/scripts/serve-frank-large.sh index d7d42a4..d74f6bf 100755 --- a/scripts/serve-frank-large.sh +++ b/scripts/serve-frank-large.sh @@ -43,7 +43,7 @@ fi : "${UBATCH_SIZE:=2048}" : "${CACHE_TYPE_K:=f16}" : "${CACHE_TYPE_V:=f16}" -: "${CTX_CHECKPOINTS:=32}" +: "${CTX_CHECKPOINTS:=16}" : "${CACHE_RAM:=8192}" : "${SPEC_DRAFT_N_MAX:=4}" : "${SPEC_DRAFT_N_MIN:=0}"