diff --git a/examples/aishell/asr0/local/train.sh b/examples/aishell/asr0/local/train.sh
index 54c642b6..6cf6855d 100755
--- a/examples/aishell/asr0/local/train.sh
+++ b/examples/aishell/asr0/local/train.sh
@@ -20,8 +20,7 @@ if [ ${seed} != 0 ]; then
     export FLAGS_cudnn_deterministic=True
 fi
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/aishell/asr1/local/train.sh b/examples/aishell/asr1/local/train.sh
index 1c8593bd..643a7b68 100755
--- a/examples/aishell/asr1/local/train.sh
+++ b/examples/aishell/asr1/local/train.sh
@@ -27,9 +27,8 @@ ckpt_name=$2
 
 mkdir -p exp
 
-python3 -u ${BIN_DIR}/train.py \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --seed ${seed} \
---ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --profiler-options "${profiler_options}" \
diff --git a/examples/callcenter/asr1/local/train.sh b/examples/callcenter/asr1/local/train.sh
index 3e92fd16..e919e0d2 100755
--- a/examples/callcenter/asr1/local/train.sh
+++ b/examples/callcenter/asr1/local/train.sh
@@ -21,8 +21,7 @@ if [ ${seed} != 0 ]; then
     export FLAGS_cudnn_deterministic=True
 fi
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/librispeech/asr0/local/train.sh b/examples/librispeech/asr0/local/train.sh
index 0479398f..55130fc1 100755
--- a/examples/librispeech/asr0/local/train.sh
+++ b/examples/librispeech/asr0/local/train.sh
@@ -20,8 +20,7 @@ if [ ${seed} != 0 ]; then
     export FLAGS_cudnn_deterministic=True
 fi
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/librispeech/asr1/local/train.sh b/examples/librispeech/asr1/local/train.sh
index 275d3a49..b3a4160e 100755
--- a/examples/librispeech/asr1/local/train.sh
+++ b/examples/librispeech/asr1/local/train.sh
@@ -22,8 +22,7 @@ fi
 # export FLAGS_cudnn_exhaustive_search=true
 # export FLAGS_conv_workspace_size_limit=4000
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/librispeech/asr2/local/train.sh b/examples/librispeech/asr2/local/train.sh
index 898391f4..06c7a0e2 100755
--- a/examples/librispeech/asr2/local/train.sh
+++ b/examples/librispeech/asr2/local/train.sh
@@ -19,9 +19,8 @@ if [ ${seed} != 0 ]; then
     export FLAGS_cudnn_deterministic=True
 fi
 
-python3 -u ${BIN_DIR}/train.py \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --model-name u2_kaldi \
---ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/timit/asr1/local/train.sh b/examples/timit/asr1/local/train.sh
index 9b3fa177..c4648c3e 100755
--- a/examples/timit/asr1/local/train.sh
+++ b/examples/timit/asr1/local/train.sh
@@ -19,8 +19,7 @@ if [ ${seed} != 0  ]; then
     export FLAGS_cudnn_deterministic=True
 fi
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/tiny/asr0/local/train.sh b/examples/tiny/asr0/local/train.sh
index a69b6ddb..33d8a238 100755
--- a/examples/tiny/asr0/local/train.sh
+++ b/examples/tiny/asr0/local/train.sh
@@ -26,8 +26,7 @@ model_type=$3
 
 mkdir -p exp
 
-python3 -u ${BIN_DIR}/train.py \
---ngpu ${ngpu} \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/tiny/asr1/local/train.sh b/examples/tiny/asr1/local/train.sh
index 1c8593bd..643a7b68 100755
--- a/examples/tiny/asr1/local/train.sh
+++ b/examples/tiny/asr1/local/train.sh
@@ -27,9 +27,8 @@ ckpt_name=$2
 
 mkdir -p exp
 
-python3 -u ${BIN_DIR}/train.py \
+python3 -m paddle.distributed.launch --gpus=${CUDA_VISIBLE_DEVICES} ${BIN_DIR}/train.py \
 --seed ${seed} \
---ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --profiler-options "${profiler_options}" \
diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/train.py b/paddlespeech/s2t/exps/deepspeech2/bin/train.py
index 09e8662f..9c7c61a5 100644
--- a/paddlespeech/s2t/exps/deepspeech2/bin/train.py
+++ b/paddlespeech/s2t/exps/deepspeech2/bin/train.py
@@ -27,10 +27,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.ngpu > 1:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
-    else:
-        main_sp(config, args)
+    main_sp(config, args)
 
 
 if __name__ == "__main__":
diff --git a/paddlespeech/s2t/exps/u2/bin/train.py b/paddlespeech/s2t/exps/u2/bin/train.py
index 53c22328..cfd7be81 100644
--- a/paddlespeech/s2t/exps/u2/bin/train.py
+++ b/paddlespeech/s2t/exps/u2/bin/train.py
@@ -32,10 +32,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.ngpu > 1:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
-    else:
-        main_sp(config, args)
+    main_sp(config, args)
 
 
 if __name__ == "__main__":
diff --git a/paddlespeech/s2t/exps/u2_kaldi/bin/train.py b/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
index fcfc05a8..f802e53d 100644
--- a/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
+++ b/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
@@ -36,10 +36,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.ngpu > 1:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
-    else:
-        main_sp(config, args)
+    main_sp(config, args)
 
 
 if __name__ == "__main__":
diff --git a/paddlespeech/s2t/exps/u2_st/bin/train.py b/paddlespeech/s2t/exps/u2_st/bin/train.py
index 4dec9ec8..8d4bfca8 100644
--- a/paddlespeech/s2t/exps/u2_st/bin/train.py
+++ b/paddlespeech/s2t/exps/u2_st/bin/train.py
@@ -30,10 +30,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.ngpu > 1:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
-    else:
-        main_sp(config, args)
+    main_sp(config, args)
 
 
 if __name__ == "__main__":