diff --git a/deepspeed/runtime/constants.py b/deepspeed/runtime/constants.py
index 9e73bad73376..d9f3ba1639e4 100755
--- a/deepspeed/runtime/constants.py
+++ b/deepspeed/runtime/constants.py
@@ -251,7 +251,7 @@
 "gradient_clipping": 1.0
 '''
 GRADIENT_CLIPPING = 'gradient_clipping'
-GRADIENT_CLIPPING_DEFAULT = 0.
+GRADIENT_CLIPPING_DEFAULT = 1.0
 
 #########################################
 # Capture graph for short kernels sequences