From 73b3de79eae5dec7459d54888083e5068890799b Mon Sep 17 00:00:00 2001
From: Xinyu Yang <cauyxy@163.com>
Date: Fri, 18 Aug 2023 03:56:04 +0800
Subject: [PATCH] explicitly del state (#784)

---
 vllm/model_executor/weight_utils.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/vllm/model_executor/weight_utils.py b/vllm/model_executor/weight_utils.py
index a9d899ad..3127a360 100644
--- a/vllm/model_executor/weight_utils.py
+++ b/vllm/model_executor/weight_utils.py
@@ -76,6 +76,8 @@ def hf_model_weights_iterator(
             state = torch.load(bin_file, map_location="cpu")
             for name, param in state.items():
                 yield name, param
+            del state
+            torch.cuda.empty_cache()
 
 
 def load_tensor_parallel_weights(