update README: reorder FlagOS and model download steps

README.md

@@ -60,18 +60,18 @@ The MAPE between CUDA and FlagOS(CUDA as ground truth) is 2.2994%. You can easil
 
 ## Operation Steps
 
-### Download Open-source Model Weights
+### Download FlagOS Image
 
 ```bash
-pip install modelscope
-modelscope download --model BAAI/RoboBrain-X0-Preview --local_dir /share/RoboBrain-X0-Preview
-
+docker pull harbor.baai.ac.cn/flagrelease-public/flagrelease_nvidia_x0_norand
 ```
 
-### Download FlagOS Image
+### Download Open-source Model Weights
 
 ```bash
-docker pull harbor.baai.ac.cn/flagrelease-public/flagrelease_nvidia_x0_norand
+pip install modelscope
+modelscope download --model BAAI/RoboBrain-X0-Preview --local_dir /share/RoboBrain-X0-Preview
+
 ```
 
 ### Start the inference service
@@ -133,7 +133,6 @@ torch.backends.cudnn.benchmark = False
 
 1. You can try to load Model in BF16 for lower memory occupancy or quicker inference. But BF16 have only 7 precision bits, which cannnot constrain MAPE under 5% even if you launch two CUDA server on different GPUid on the same GPU server. 
 
-
 # Contributing
 
 We warmly welcome global developers to join us:
@@ -143,7 +142,6 @@ We warmly welcome global developers to join us:
 3. Improve technical documentation
 4. Expand hardware adaptation support
 
-
 # License
 
 本模型的权重来源于BAAI/RoboBrain-X0-Preview，以apache2.0协议https://www.apache.org/licenses/LICENSE-2.0.txt开源。