From 948114e3e3325af01c31697dee861459624ca805 Mon Sep 17 00:00:00 2001
From: Mark Ward <mark.ward@binaryward.com>
Date: Sun, 28 Apr 2024 16:41:38 +0000
Subject: [PATCH] fix sched to wait for the runner to terminate to ensure
 following vram check will be more accurate

---
 llm/server.go | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/llm/server.go b/llm/server.go
index 231100a0..c82c5099 100644
--- a/llm/server.go
+++ b/llm/server.go
@@ -899,7 +899,10 @@ func (s *llmServer) Detokenize(ctx context.Context, tokens []int) (string, error
 func (s *llmServer) Close() error {
 	if s.cmd != nil {
 		slog.Debug("stopping llama server")
-		return s.cmd.Process.Kill()
+		if err := s.cmd.Process.Kill(); err != nil {
+			return err
+		}
+		return s.cmd.Wait()
 	}
 
 	return nil