Add preload env in run.sh to fix xft-vllm perf issue (#204)

* add preload env in run.sh Signed-off-by: lvliang-intel <[email protected]> * update comments Signed-off-by: lvliang-intel <[email protected]> --------- Signed-off-by: lvliang-intel <[email protected]>
opea-project · Jun 21, 2024 · 309c2da · 309c2da
1 parent 7589142
commit 309c2da
Showing 1 changed file with 3 additions and 1 deletion.
diff --git a/comps/llms/text-generation/vllm-xft/run.sh b/comps/llms/text-generation/vllm-xft/run.sh
@@ -3,7 +3,9 @@
 # Copyright (C) 2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
-# Preload libiomp5.so by following cmd or LD_PRELOAD=libiomp5.so manually
+# Preloading libiomp5.so is essential for optimal performance.
+# libiomp5.so is the Intel OpenMP runtime library, providing parallel computation support,
+# thread management, task scheduling, and performance optimization on Intel X86 platforms.
 export $(python -c 'import xfastertransformer as xft; print(xft.get_env())')
 
 # convert the model to fastertransformer format