From ed5bc0068e8de846ca2ef8a1b2afcb144529df31 Mon Sep 17 00:00:00 2001
From: hh-space-invader <h.hagag.ali@gmail.com>
Date: Fri, 28 Feb 2025 10:55:18 +0200
Subject: [PATCH] fix: Minimize gpu memory fragmentation

---
 fastembed/common/onnx_model.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fastembed/common/onnx_model.py b/fastembed/common/onnx_model.py
index 52b08b43..3bb7b8ec 100644
--- a/fastembed/common/onnx_model.py
+++ b/fastembed/common/onnx_model.py
@@ -85,7 +85,7 @@ def _load_onnx_model(
 
         so = ort.SessionOptions()
         so.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL
-
+        so.add_session_config_entry("memory.enable_memory_arena_shrinkage", "1")
         if threads is not None:
             so.intra_op_num_threads = threads
             so.inter_op_num_threads = threads