diff --git a/docs/en/get_started/installation.md b/docs/en/get_started/installation.md
index b3e8bb8abd..c00111c2ab 100644
--- a/docs/en/get_started/installation.md
+++ b/docs/en/get_started/installation.md
@@ -23,7 +23,7 @@ pip install lmdeploy
 The default prebuilt package is compiled on **CUDA 12**. If CUDA 11+ (>=11.3) is required, you can install lmdeploy by:
 
 ```shell
-export LMDEPLOY_VERSION=0.6.3
+export LMDEPLOY_VERSION=0.6.4
 export PYTHON_VERSION=38
 pip install https://github.com/InternLM/lmdeploy/releases/download/v${LMDEPLOY_VERSION}/lmdeploy-${LMDEPLOY_VERSION}+cu118-cp${PYTHON_VERSION}-cp${PYTHON_VERSION}-manylinux2014_x86_64.whl --extra-index-url https://download.pytorch.org/whl/cu118
 ```
diff --git a/docs/zh_cn/get_started/installation.md b/docs/zh_cn/get_started/installation.md
index 12562c51d5..0213fa6d15 100644
--- a/docs/zh_cn/get_started/installation.md
+++ b/docs/zh_cn/get_started/installation.md
@@ -23,7 +23,7 @@ pip install lmdeploy
 默认的预构建包是在 **CUDA 12** 上编译的。如果需要 CUDA 11+ (>=11.3)，你可以使用以下命令安装 lmdeploy：
 
 ```shell
-export LMDEPLOY_VERSION=0.6.3
+export LMDEPLOY_VERSION=0.6.4
 export PYTHON_VERSION=38
 pip install https://github.com/InternLM/lmdeploy/releases/download/v${LMDEPLOY_VERSION}/lmdeploy-${LMDEPLOY_VERSION}+cu118-cp${PYTHON_VERSION}-cp${PYTHON_VERSION}-manylinux2014_x86_64.whl --extra-index-url https://download.pytorch.org/whl/cu118
 ```
diff --git a/lmdeploy/pytorch/models/patch.py b/lmdeploy/pytorch/models/patch.py
index 9da1b9f4ea..a7fe4431ed 100644
--- a/lmdeploy/pytorch/models/patch.py
+++ b/lmdeploy/pytorch/models/patch.py
@@ -8,6 +8,7 @@
 
 import torch
 from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_utils import load_state_dict
 
 from lmdeploy.utils import get_logger
 
@@ -295,7 +296,9 @@ def add_adapters(model: torch.nn.Module,
     for name, path in adapters.items():
         adapter_id = adapter_id_map[name]
         checkpoint_path = f'{path}/adapter_model.bin'
-        state_dict = torch.load(checkpoint_path, map_location=device)
+        if not osp.exists(checkpoint_path):
+            checkpoint_path = f'{path}/adapter_model.safetensors'
+        state_dict = load_state_dict(checkpoint_path, map_location=device)
 
         if hasattr(model, 'load_lora_weights'):
             model.load_lora_weights(state_dict.items(), adapter_id=adapter_id)
diff --git a/lmdeploy/version.py b/lmdeploy/version.py
index d9f4307a78..f705fcb332 100644
--- a/lmdeploy/version.py
+++ b/lmdeploy/version.py
@@ -1,7 +1,7 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 from typing import Tuple
 
-__version__ = '0.6.3'
+__version__ = '0.6.4'
 short_version = __version__
 
 
diff --git a/requirements/runtime_ascend.txt b/requirements/runtime_ascend.txt
index 05d74bbe72..c5d44cc995 100644
--- a/requirements/runtime_ascend.txt
+++ b/requirements/runtime_ascend.txt
@@ -1,5 +1,5 @@
 accelerate>=0.29.3
-dlinfer-ascend>=0.1.2
+dlinfer-ascend>=0.1.3
 einops
 fastapi
 fire