diff --git a/gnn/GCN/paddlepaddle/README.md b/gnn/GCN/paddlepaddle/README.md
index b44346cc90504d83647872ad38529f17a75e2c31..c713af33ed8ac232f690808366e2ee31b42e9668 100644
--- a/gnn/GCN/paddlepaddle/README.md
+++ b/gnn/GCN/paddlepaddle/README.md
@@ -15,9 +15,9 @@ git clone https://github.com/PaddlePaddle/PGL.git
 
 ```
 # Pip the requirements
-pip install pgl
-pip install urllib3==1.23
-pip install networkx
+pip3 install pgl
+pip3 install urllib3==1.23
+pip3 install networkx
 ```
 
 ## Step 2:Preparing datasets
@@ -35,16 +35,22 @@ cd PGL/examples/gcn/
 CUDA_VISIBLE_DEVICES=0 python3 train.py --dataset cora
 
 # Run on CPU 
-CUDA_VISIBLE_DEVICES= python3 train.py --dataset cora
+python3 train.py --dataset cora
 ```
 
 ## Results
+| GPUS       | Datasets | speed    |
+| ---------- | -------- | -------- |
+| BI V100×1 | CORA     | 0.0064   |
+| BI V100×1 | PUBMED   | 0.0076   |
+| BI V100×1 | CITESEER | 0.0085   |
+
 
 | GPUS       | Datasets | Accurary |
 | ---------- | -------- | -------- |
 | BI V100×1 | CORA     | 80.3%    |
-| BI V100×1 | PUBMED   | 79.3%    |
-| BI V100×1 | CITESEER | 71.0%    |
+| BI V100×1 | PUBMED   | 79.0%    |
+| BI V100×1 | CITESEER | 70.6%    |
 
 ## Reference