diff --git a/benchmark/IntelOptimizedPaddle.md b/benchmark/IntelOptimizedPaddle.md
index 1bf9ea9df02a1f0e0b71400207a9f375a2b3d25b..040f5ffa41968cbf93a817faa1db86c18956341e 100644
--- a/benchmark/IntelOptimizedPaddle.md
+++ b/benchmark/IntelOptimizedPaddle.md
@@ -23,7 +23,7 @@ On each machine, we will test and compare the performance of training on single
 ## Benchmark Model
 
 ### Server
-Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148M CPU @ 2.40GHz
+Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
 
 Input image size - 3 * 224 * 224, Time: images/second
 
diff --git a/python/paddle/v2/plot/plot.py b/python/paddle/v2/plot/plot.py
index 6f7bd039b07db4832295c2374293bffa588eb4ef..c18e63dd5f60481ba804738a6a9238dfea35d9f3 100644
--- a/python/paddle/v2/plot/plot.py
+++ b/python/paddle/v2/plot/plot.py
@@ -56,7 +56,7 @@ class Ploter(object):
         assert isinstance(data, PlotData)
         data.append(step, value)
 
-    def plot(self):
+    def plot(self, path=None):
         if self.__plot_is_disabled__():
             return
 
@@ -68,8 +68,11 @@ class Ploter(object):
                 titles.append(title)
                 self.plt.plot(data.step, data.value)
         self.plt.legend(titles, loc='upper left')
-        self.display.clear_output(wait=True)
-        self.display.display(self.plt.gcf())
+        if path is None:
+            self.display.clear_output(wait=True)
+            self.display.display(self.plt.gcf())
+        else:
+            self.plt.savefig(path)
         self.plt.gcf().clear()
 
     def reset(self):