diff --git a/run_classifier_pytorch.py b/run_classifier_pytorch.py
index c8ec8ab6e23..3410a09b045 100644
--- a/run_classifier_pytorch.py
+++ b/run_classifier_pytorch.py
@@ -427,7 +427,10 @@ def main():
                         type=int,
                         default=-1,
                         help="local_rank for distributed training on gpus")
-
+    parser.add_argument('--seed', 
+                        type=int, 
+                        default=42,
+                        help="random seed for initialization")
     args = parser.parse_args()
 
     processors = {
@@ -444,7 +447,12 @@ def main():
         n_gpu = 1
         # print("Initializing the distributed backend: NCCL")
     print("device", device, "n_gpu", n_gpu)
-
+    
+    random.seed(args.seed)
+    np.random.seed(args.seed)
+    torch.manual_seed(args.seed)
+    if n_gpu>0: torch.cuda.manual_seed_all(args.seed)
+    
     if not args.do_train and not args.do_eval:
         raise ValueError("At least one of `do_train` or `do_eval` must be True.")
 
diff --git a/run_squad_pytorch.py b/run_squad_pytorch.py
index 2a67262d96e..a1db682cd4f 100644
--- a/run_squad_pytorch.py
+++ b/run_squad_pytorch.py
@@ -745,6 +745,10 @@ def main():
                         type=int,
                         default=-1,
                         help="local_rank for distributed training on gpus")
+    parser.add_argument('--seed', 
+                    type=int, 
+                    default=42,
+                    help="random seed for initialization")
 
     args = parser.parse_args()
 
@@ -757,6 +761,11 @@ def main():
         # print("Initializing the distributed backend: NCCL")
     print("device", device, "n_gpu", n_gpu)
 
+    random.seed(args.seed)
+    np.random.seed(args.seed)
+    torch.manual_seed(args.seed)
+    if n_gpu>0: torch.cuda.manual_seed_all(args.seed)
+
     if not args.do_train and not args.do_predict:
         raise ValueError("At least one of `do_train` or `do_predict` must be True.")