use all 32k context for mixtral llama.cpp daemon

groovybits · Mar 27, 2024 · fa5d739 · fa5d739
1 parent bd31091
commit fa5d739
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/bin/llama_start.sh b/bin/llama_start.sh
@@ -6,7 +6,7 @@ MODEL=/Volumes/BrahmaSSD/LLM/models/GGUF/dolphin-2.7-mixtral-8x7b.Q8_0.gguf
 server \
     -m $MODEL \
     -c 0 \
-    -np 2 \
+    -np 1\
     --port 8080 \
     -ngl 60 \
     -t 24 \