diff --git a/model_examples/OpenDWM/README.md b/model_examples/OpenDWM/README.md
index ce49c54a599d54b7130eb13975fcb830693c7ae6..a1ca7d66387e26c963bf845a78416dfa476842b1 100644
--- a/model_examples/OpenDWM/README.md
+++ b/model_examples/OpenDWM/README.md
@@ -129,10 +129,10 @@ OpenDWM是一种统一的多视角驾驶视频生成框架。通过融合单/多
     # 安装其他依赖项
     python -m pip install -r requirements.txt
 
-    # 安装tbe和hccl
+    # 安装tbe和hccl, 将 CANN 包目录记作 cann_root_dir
     pip uninstall te topi hccl -y
-    pip install $ASCEND_CUSTOM_PATH/latest/lib64/te-*-py3-none-any.whl
-    pip install $ASCEND_CUSTOM_PATH/latest/lib64/hccl-*-py3-none-any.whl
+    pip install {cann_root_dir}/latest/lib64/te-*-py3-none-any.whl
+    pip install {cann_root_dir}/latest/lib64/hccl-*-py3-none-any.whl
     ```
     
 6. 使用 patch 文件：
@@ -281,4 +281,6 @@ apt-get update && apt-get install -y gawk
 yum install -y gawk
 ```
 
-2. 训练过程会自动下载inception权重，如果遇到网络问题等下载失败，可以本地下载后，手动将该权重文件放到日志指定路径
\ No newline at end of file
+2. 训练过程会自动下载inception权重，如果遇到网络问题等下载失败，可以本地下载后，手动将该权重文件放到日志指定路径
+
+3. 我们支持的训练方式是单数据集nuScenes，若在执行`python -m pip install -r requirements.txt` 时由于网络原因下载kitti仓失败，注释掉对应依赖即可
\ No newline at end of file
diff --git a/model_examples/OpenDWM/test/train.sh b/model_examples/OpenDWM/test/train.sh
index d1ba6cc2d2abea6986645df0a5454c127dc9d2ab..118d72e298043a40f7a4fe9038ae93be39c2cac4 100644
--- a/model_examples/OpenDWM/test/train.sh
+++ b/model_examples/OpenDWM/test/train.sh
@@ -37,5 +37,5 @@ python -m torch.distributed.run \
     --node-rank $NODE_RANK \
     src/dwm/train.py \
     -c $CONFIG_FILE \
-    -o $OUTPUT_DIR 2>&1 | tee "${OUTPUT_DIR}/train.log"
+    -o $OUTPUT_DIR > "${OUTPUT_DIR}/train.log" 2>&1
     
\ No newline at end of file
diff --git a/model_examples/OpenDWM/test/train_performance.sh b/model_examples/OpenDWM/test/train_performance.sh
index 5221d69aa84055518c298c522fa1b7d1057e00c5..47c59de0a0923b9779086a16a0deb6d04a00c8d9 100644
--- a/model_examples/OpenDWM/test/train_performance.sh
+++ b/model_examples/OpenDWM/test/train_performance.sh
@@ -38,7 +38,7 @@ python -m torch.distributed.run \
     --node-rank $NODE_RANK \
     src/dwm/train.py \
     -c $CONFIG_FILE \
-    -o $OUTPUT_DIR 2>&1 | tee "${OUTPUT_DIR}/train_performance.log"
+    -o $OUTPUT_DIR > "${OUTPUT_DIR}/train_performance.log" 2>&1
 
 LOG_FILE="${OUTPUT_DIR}/train_performance.log"