diff --git a/examples/run_classifier.py b/examples/run_classifier.py
index ccb25bfc..e1dcd363 100644
--- a/examples/run_classifier.py
+++ b/examples/run_classifier.py
@@ -558,8 +558,6 @@ def main():
     # Load a trained model that you have fine-tuned
     model_state_dict = torch.load(output_model_file)
     model = BertForSequenceClassification.from_pretrained(args.bert_model, state_dict=model_state_dict)
-    if args.fp16:
-        model.half()
     model.to(device)
 
     if args.do_eval and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
diff --git a/examples/run_squad.py b/examples/run_squad.py
index fb8cd8dd..d6e96f4a 100644
--- a/examples/run_squad.py
+++ b/examples/run_squad.py
@@ -923,8 +923,6 @@ def main():
     # Load a trained model that you have fine-tuned
     model_state_dict = torch.load(output_model_file)
     model = BertForQuestionAnswering.from_pretrained(args.bert_model, state_dict=model_state_dict)
-    if args.fp16:
-        model.half()
     model.to(device)
 
     if args.do_predict and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
diff --git a/examples/run_swag.py b/examples/run_swag.py
index 8d2ab40f..bedfff0b 100644
--- a/examples/run_swag.py
+++ b/examples/run_swag.py
@@ -478,8 +478,6 @@ def main():
     model = BertForMultipleChoice.from_pretrained(args.bert_model,
         state_dict=model_state_dict,
         num_choices=4)
-    if args.fp16:
-        model.half()
     model.to(device)
 
     if args.do_eval and (args.local_rank == -1 or torch.distributed.get_rank() == 0):