From 9352eeb752531decccc7c6b91a07bc3dd5efa67e Mon Sep 17 00:00:00 2001
From: Richard Lyons <frob@cloudstaff.com>
Date: Sun, 18 Aug 2024 02:55:01 +0200
Subject: [PATCH] Reset NumCtx.

---
 server/sched.go | 1 +
 1 file changed, 1 insertion(+)

diff --git a/server/sched.go b/server/sched.go
index 4d9c0296..3fe6d7fc 100644
--- a/server/sched.go
+++ b/server/sched.go
@@ -736,6 +736,7 @@ func pickBestFullFitByLibrary(req *LlmRequest, ggml *llm.GGML, gpus gpu.GpuInfoL
 func pickBestPartialFitByLibrary(req *LlmRequest, ggml *llm.GGML, gpus gpu.GpuInfoList, numParallel *int) gpu.GpuInfoList {
 	if *numParallel <= 0 {
 		*numParallel = 1
+                req.opts.NumCtx = req.origNumCtx
         }
 	byLibrary := gpus.ByLibrary()
 	if len(byLibrary) <= 1 {