diff --git a/docs/source/tts/advanced_usage.md b/docs/source/tts/advanced_usage.md
index 0540a1c3..04088964 100644
--- a/docs/source/tts/advanced_usage.md
+++ b/docs/source/tts/advanced_usage.md
@@ -290,7 +290,7 @@ The following is the basic  `ArgumentParser`:
 1. `--config`  is used to support configuration file parsing, and the configuration file itself handles the unique options of each experiment.
 2. `--train-metadata` is the path to the training data.
 3.  `--output-dir` is the dir to save the training results.（if there are checkpoints in  `checkpoints/` of  `--output-dir` , it's defalut to reload the newest checkpoint to train)
-4. `--device` and  `--nprocs` determine operation modes，`--device` specifies the type of running device, whether to run on `cpu` or `gpu`. `--nprocs` refers to  the number of training processes. If `nprocs` > 1, it means that multi process parallel training is used. (Note: currently only GPU multi card multi process training is supported.)
+4. `--ngpu` determine operation modes，`--ngpu` refers to the number of training processes. If `ngpu` > 0, it means using GPU, else CPU is used.
 
 Developers can refer to the examples in `examples` to write the default configuration file when adding new experiments.
 
diff --git a/examples/aishell/asr0/local/export.sh b/examples/aishell/asr0/local/export.sh
index a5e62c28..426a72fe 100755
--- a/examples/aishell/asr0/local/export.sh
+++ b/examples/aishell/asr0/local/export.sh
@@ -14,7 +14,7 @@ jit_model_export_path=$3
 model_type=$4
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path} \
diff --git a/examples/aishell/asr0/local/test.sh b/examples/aishell/asr0/local/test.sh
index 2ae0740b..8cbff235 100755
--- a/examples/aishell/asr0/local/test.sh
+++ b/examples/aishell/asr0/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/aishell/asr0/local/test_export.sh b/examples/aishell/asr0/local/test_export.sh
index f0a30ce5..4f5e5c8b 100755
--- a/examples/aishell/asr0/local/test_export.sh
+++ b/examples/aishell/asr0/local/test_export.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test_export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${jit_model_export_path}.rsl \
 --export_path ${jit_model_export_path} \
diff --git a/examples/aishell/asr0/local/test_hub.sh b/examples/aishell/asr0/local/test_hub.sh
index d01496c4..b9cb7fa0 100755
--- a/examples/aishell/asr0/local/test_hub.sh
+++ b/examples/aishell/asr0/local/test_hub.sh
@@ -20,7 +20,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test_hub.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/aishell/asr0/local/train.sh b/examples/aishell/asr0/local/train.sh
index edbf3383..54c642b6 100755
--- a/examples/aishell/asr0/local/train.sh
+++ b/examples/aishell/asr0/local/train.sh
@@ -21,7 +21,7 @@ if [ ${seed} != 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/aishell/asr1/local/align.sh b/examples/aishell/asr1/local/align.sh
index 279461aa..c65d611c 100755
--- a/examples/aishell/asr1/local/align.sh
+++ b/examples/aishell/asr1/local/align.sh
@@ -18,7 +18,7 @@ mkdir -p ${output_dir}
 # align dump in `result_file`
 # .tier, .TextGrid dump in `dir of result_file`
 python3 -u ${BIN_DIR}/alignment.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/aishell/asr1/local/export.sh b/examples/aishell/asr1/local/export.sh
index b562218e..6b646b46 100755
--- a/examples/aishell/asr1/local/export.sh
+++ b/examples/aishell/asr1/local/export.sh
@@ -13,7 +13,7 @@ ckpt_path_prefix=$2
 jit_model_export_path=$3
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/aishell/asr1/local/test.sh b/examples/aishell/asr1/local/test.sh
index 47bd2f63..da159de7 100755
--- a/examples/aishell/asr1/local/test.sh
+++ b/examples/aishell/asr1/local/test.sh
@@ -34,7 +34,7 @@ for type in attention ctc_greedy_search; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
@@ -53,7 +53,7 @@ for type in ctc_prefix_beam_search attention_rescoring; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/aishell/asr1/local/test_hub.sh b/examples/aishell/asr1/local/test_hub.sh
index 6e78ec78..0fd30901 100755
--- a/examples/aishell/asr1/local/test_hub.sh
+++ b/examples/aishell/asr1/local/test_hub.sh
@@ -29,7 +29,7 @@ for type in  attention_rescoring; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test_hub.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/aishell/asr1/local/train.sh b/examples/aishell/asr1/local/train.sh
index 71af3a00..1c8593bd 100755
--- a/examples/aishell/asr1/local/train.sh
+++ b/examples/aishell/asr1/local/train.sh
@@ -29,7 +29,7 @@ mkdir -p exp
 
 python3 -u ${BIN_DIR}/train.py \
 --seed ${seed} \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --profiler-options "${profiler_options}" \
diff --git a/examples/callcenter/asr1/local/align.sh b/examples/callcenter/asr1/local/align.sh
index b679e2ea..681c77ed 100755
--- a/examples/callcenter/asr1/local/align.sh
+++ b/examples/callcenter/asr1/local/align.sh
@@ -23,7 +23,7 @@ mkdir -p ${output_dir}
 # align dump in `result_file`
 # .tier, .TextGrid dump in `dir of result_file`
 python3 -u ${BIN_DIR}/alignment.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/callcenter/asr1/local/export.sh b/examples/callcenter/asr1/local/export.sh
index d5f912e9..36de2d66 100755
--- a/examples/callcenter/asr1/local/export.sh
+++ b/examples/callcenter/asr1/local/export.sh
@@ -13,7 +13,7 @@ ckpt_path_prefix=$2
 jit_model_export_path=$3
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/callcenter/asr1/local/test.sh b/examples/callcenter/asr1/local/test.sh
index 0aa99e19..fc43c5a2 100755
--- a/examples/callcenter/asr1/local/test.sh
+++ b/examples/callcenter/asr1/local/test.sh
@@ -28,7 +28,7 @@ for type in attention ctc_greedy_search; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
@@ -47,7 +47,7 @@ for type in ctc_prefix_beam_search attention_rescoring; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/callcenter/asr1/local/train.sh b/examples/callcenter/asr1/local/train.sh
index eb8f8662..3e92fd16 100755
--- a/examples/callcenter/asr1/local/train.sh
+++ b/examples/callcenter/asr1/local/train.sh
@@ -22,7 +22,7 @@ if [ ${seed} != 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/librispeech/asr0/local/export.sh b/examples/librispeech/asr0/local/export.sh
index a5e62c28..426a72fe 100755
--- a/examples/librispeech/asr0/local/export.sh
+++ b/examples/librispeech/asr0/local/export.sh
@@ -14,7 +14,7 @@ jit_model_export_path=$3
 model_type=$4
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path} \
diff --git a/examples/librispeech/asr0/local/test.sh b/examples/librispeech/asr0/local/test.sh
index 4d00f30b..a627ef72 100755
--- a/examples/librispeech/asr0/local/test.sh
+++ b/examples/librispeech/asr0/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr0/local/test_hub.sh b/examples/librispeech/asr0/local/test_hub.sh
index 2e32f24a..fd9a603a 100755
--- a/examples/librispeech/asr0/local/test_hub.sh
+++ b/examples/librispeech/asr0/local/test_hub.sh
@@ -20,7 +20,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test_hub.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr0/local/train.sh b/examples/librispeech/asr0/local/train.sh
index 519df7fe..0479398f 100755
--- a/examples/librispeech/asr0/local/train.sh
+++ b/examples/librispeech/asr0/local/train.sh
@@ -21,7 +21,7 @@ if [ ${seed} != 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/librispeech/asr1/local/align.sh b/examples/librispeech/asr1/local/align.sh
index 279461aa..c65d611c 100755
--- a/examples/librispeech/asr1/local/align.sh
+++ b/examples/librispeech/asr1/local/align.sh
@@ -18,7 +18,7 @@ mkdir -p ${output_dir}
 # align dump in `result_file`
 # .tier, .TextGrid dump in `dir of result_file`
 python3 -u ${BIN_DIR}/alignment.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr1/local/export.sh b/examples/librispeech/asr1/local/export.sh
index b562218e..6b646b46 100755
--- a/examples/librispeech/asr1/local/export.sh
+++ b/examples/librispeech/asr1/local/export.sh
@@ -13,7 +13,7 @@ ckpt_path_prefix=$2
 jit_model_export_path=$3
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/librispeech/asr1/local/test.sh b/examples/librispeech/asr1/local/test.sh
index ceaa77cf..aa06132e 100755
--- a/examples/librispeech/asr1/local/test.sh
+++ b/examples/librispeech/asr1/local/test.sh
@@ -50,7 +50,7 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
             batch_size=64
         fi
         python3 -u ${BIN_DIR}/test.py \
-            --nproc ${ngpu} \
+            --ngpu ${ngpu} \
             --config ${config_path} \
             --result_file ${ckpt_prefix}.${type}.rsl \
             --checkpoint_path ${ckpt_prefix} \
@@ -74,7 +74,7 @@ for type in ctc_greedy_search; do
         batch_size=64
     fi
     python3 -u ${BIN_DIR}/test.py \
-        --nproc ${ngpu} \
+        --ngpu ${ngpu} \
         --config ${config_path} \
         --result_file ${ckpt_prefix}.${type}.rsl \
         --checkpoint_path ${ckpt_prefix} \
@@ -94,7 +94,7 @@ for type in ctc_prefix_beam_search attention_rescoring; do
     echo "decoding ${type}"
     batch_size=1
     python3 -u ${BIN_DIR}/test.py \
-        --nproc ${ngpu} \
+        --ngpu ${ngpu} \
         --config ${config_path} \
         --result_file ${ckpt_prefix}.${type}.rsl \
         --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr1/local/test_hub.sh b/examples/librispeech/asr1/local/test_hub.sh
index dcf242e9..46bd8bc2 100755
--- a/examples/librispeech/asr1/local/test_hub.sh
+++ b/examples/librispeech/asr1/local/test_hub.sh
@@ -36,7 +36,7 @@ for type in attention_rescoring; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test_hub.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr1/local/train.sh b/examples/librispeech/asr1/local/train.sh
index 8f92c646..275d3a49 100755
--- a/examples/librispeech/asr1/local/train.sh
+++ b/examples/librispeech/asr1/local/train.sh
@@ -23,7 +23,7 @@ fi
 # export FLAGS_conv_workspace_size_limit=4000
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/librispeech/asr2/local/align.sh b/examples/librispeech/asr2/local/align.sh
index b45f4a0f..626c3574 100755
--- a/examples/librispeech/asr2/local/align.sh
+++ b/examples/librispeech/asr2/local/align.sh
@@ -22,7 +22,7 @@ python3 -u ${BIN_DIR}/test.py \
 --model-name 'u2_kaldi' \
 --run-mode 'align' \
 --dict-path ${dict_path} \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result-file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr2/local/export.sh b/examples/librispeech/asr2/local/export.sh
index 9c66dc62..1bdce16c 100755
--- a/examples/librispeech/asr2/local/export.sh
+++ b/examples/librispeech/asr2/local/export.sh
@@ -15,7 +15,7 @@ jit_model_export_path=$3
 python3 -u ${BIN_DIR}/test.py \
 --model-name 'u2_kaldi' \
 --run-mode 'export' \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/librispeech/asr2/local/test.sh b/examples/librispeech/asr2/local/test.sh
index 23670f74..d210f2a8 100755
--- a/examples/librispeech/asr2/local/test.sh
+++ b/examples/librispeech/asr2/local/test.sh
@@ -76,7 +76,7 @@ for dmethd in attention ctc_greedy_search ctc_prefix_beam_search attention_resco
             python3 -u ${BIN_DIR}/test.py \
             --model-name u2_kaldi \
             --run-mode test \
-            --nproc ${ngpu} \
+            --ngpu ${ngpu} \
             --dict-path ${dict} \
             --config ${config_path} \
             --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/librispeech/asr2/local/train.sh b/examples/librispeech/asr2/local/train.sh
index 33b46c20..898391f4 100755
--- a/examples/librispeech/asr2/local/train.sh
+++ b/examples/librispeech/asr2/local/train.sh
@@ -21,7 +21,7 @@ fi
 
 python3 -u ${BIN_DIR}/train.py \
 --model-name u2_kaldi \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/other/1xt2x/aishell/local/test.sh b/examples/other/1xt2x/aishell/local/test.sh
index 2ae0740b..8cbff235 100755
--- a/examples/other/1xt2x/aishell/local/test.sh
+++ b/examples/other/1xt2x/aishell/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/other/1xt2x/baidu_en8k/local/test.sh b/examples/other/1xt2x/baidu_en8k/local/test.sh
index 4d00f30b..a627ef72 100755
--- a/examples/other/1xt2x/baidu_en8k/local/test.sh
+++ b/examples/other/1xt2x/baidu_en8k/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/other/1xt2x/librispeech/local/test.sh b/examples/other/1xt2x/librispeech/local/test.sh
index 4d00f30b..a627ef72 100755
--- a/examples/other/1xt2x/librispeech/local/test.sh
+++ b/examples/other/1xt2x/librispeech/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/other/1xt2x/src_deepspeech2x/test_model.py b/examples/other/1xt2x/src_deepspeech2x/test_model.py
index a9afc631..82e190d8 100644
--- a/examples/other/1xt2x/src_deepspeech2x/test_model.py
+++ b/examples/other/1xt2x/src_deepspeech2x/test_model.py
@@ -403,7 +403,7 @@ class DeepSpeech2Tester(DeepSpeech2Trainer):
     def setup(self):
         """Setup the experiment.
         """
-        paddle.set_device('gpu' if self.args.nprocs > 0 else 'cpu')
+        paddle.set_device('gpu' if self.args.ngpu > 0 else 'cpu')
 
         self.setup_output_dir()
         self.setup_checkpointer()
diff --git a/examples/ted_en_zh/st0/local/test.sh b/examples/ted_en_zh/st0/local/test.sh
index 7235c6f9..a9b18dd9 100755
--- a/examples/ted_en_zh/st0/local/test.sh
+++ b/examples/ted_en_zh/st0/local/test.sh
@@ -15,7 +15,7 @@ for type in fullsentence; do
     echo "decoding ${type}"
     batch_size=32
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${ckpt_prefix}.${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/ted_en_zh/st0/local/train.sh b/examples/ted_en_zh/st0/local/train.sh
index e5fd19dd..e366376b 100755
--- a/examples/ted_en_zh/st0/local/train.sh
+++ b/examples/ted_en_zh/st0/local/train.sh
@@ -20,7 +20,7 @@ if [ ${seed} != 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/ted_en_zh/st1/local/test.sh b/examples/ted_en_zh/st1/local/test.sh
index 7235c6f9..a9b18dd9 100755
--- a/examples/ted_en_zh/st1/local/test.sh
+++ b/examples/ted_en_zh/st1/local/test.sh
@@ -15,7 +15,7 @@ for type in fullsentence; do
     echo "decoding ${type}"
     batch_size=32
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${ckpt_prefix}.${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/ted_en_zh/st1/local/train_finetune.sh b/examples/ted_en_zh/st1/local/train_finetune.sh
index 36701121..e54c7fff 100755
--- a/examples/ted_en_zh/st1/local/train_finetune.sh
+++ b/examples/ted_en_zh/st1/local/train_finetune.sh
@@ -21,7 +21,7 @@ if [ ${seed} != 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --checkpoint_path ${ckpt_path} \
diff --git a/examples/timit/asr1/local/align.sh b/examples/timit/asr1/local/align.sh
index 279461aa..c65d611c 100755
--- a/examples/timit/asr1/local/align.sh
+++ b/examples/timit/asr1/local/align.sh
@@ -18,7 +18,7 @@ mkdir -p ${output_dir}
 # align dump in `result_file`
 # .tier, .TextGrid dump in `dir of result_file`
 python3 -u ${BIN_DIR}/alignment.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/timit/asr1/local/export.sh b/examples/timit/asr1/local/export.sh
index b562218e..6b646b46 100755
--- a/examples/timit/asr1/local/export.sh
+++ b/examples/timit/asr1/local/export.sh
@@ -13,7 +13,7 @@ ckpt_path_prefix=$2
 jit_model_export_path=$3
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/timit/asr1/local/test.sh b/examples/timit/asr1/local/test.sh
index 575bff57..08ee0e36 100755
--- a/examples/timit/asr1/local/test.sh
+++ b/examples/timit/asr1/local/test.sh
@@ -41,7 +41,7 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
             batch_size=64
         fi
         python3 -u ${BIN_DIR}/test.py \
-        --nproc ${ngpu} \
+        --ngpu ${ngpu} \
         --config ${config_path} \
         --result_file ${ckpt_prefix}.${type}.rsl \
         --checkpoint_path ${ckpt_prefix} \
@@ -61,7 +61,7 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
         echo "decoding ${type}"
         batch_size=1
         python3 -u ${BIN_DIR}/test.py \
-        --nproc ${ngpu}  \
+        --ngpu ${ngpu}  \
         --config ${config_path} \
         --result_file ${ckpt_prefix}.${type}.rsl \
         --checkpoint_path ${ckpt_prefix} \
@@ -80,7 +80,7 @@ if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
         echo "decoding ${type}"
         batch_size=1
         python3 -u ${BIN_DIR}/test.py \
-        --nproc ${ngpu}  \
+        --ngpu ${ngpu}  \
         --config ${config_path} \
         --result_file ${ckpt_prefix}.${type}.rsl \
         --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/timit/asr1/local/train.sh b/examples/timit/asr1/local/train.sh
index 89a64327..9b3fa177 100755
--- a/examples/timit/asr1/local/train.sh
+++ b/examples/timit/asr1/local/train.sh
@@ -20,7 +20,7 @@ if [ ${seed} != 0  ]; then
 fi
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --seed ${seed}
diff --git a/examples/tiny/asr0/local/export.sh b/examples/tiny/asr0/local/export.sh
index a5e62c28..426a72fe 100755
--- a/examples/tiny/asr0/local/export.sh
+++ b/examples/tiny/asr0/local/export.sh
@@ -14,7 +14,7 @@ jit_model_export_path=$3
 model_type=$4
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path} \
diff --git a/examples/tiny/asr0/local/test.sh b/examples/tiny/asr0/local/test.sh
index 4d00f30b..a627ef72 100755
--- a/examples/tiny/asr0/local/test.sh
+++ b/examples/tiny/asr0/local/test.sh
@@ -19,7 +19,7 @@ if [ $? -ne 0 ]; then
 fi
 
 python3 -u ${BIN_DIR}/test.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/tiny/asr0/local/train.sh b/examples/tiny/asr0/local/train.sh
index 5b87780a..a69b6ddb 100755
--- a/examples/tiny/asr0/local/train.sh
+++ b/examples/tiny/asr0/local/train.sh
@@ -27,7 +27,7 @@ model_type=$3
 mkdir -p exp
 
 python3 -u ${BIN_DIR}/train.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --model_type ${model_type} \
diff --git a/examples/tiny/asr1/local/align.sh b/examples/tiny/asr1/local/align.sh
index 279461aa..c65d611c 100755
--- a/examples/tiny/asr1/local/align.sh
+++ b/examples/tiny/asr1/local/align.sh
@@ -18,7 +18,7 @@ mkdir -p ${output_dir}
 # align dump in `result_file`
 # .tier, .TextGrid dump in `dir of result_file`
 python3 -u ${BIN_DIR}/alignment.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --result_file ${output_dir}/${type}.align \
 --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/tiny/asr1/local/export.sh b/examples/tiny/asr1/local/export.sh
index b562218e..6b646b46 100755
--- a/examples/tiny/asr1/local/export.sh
+++ b/examples/tiny/asr1/local/export.sh
@@ -13,7 +13,7 @@ ckpt_path_prefix=$2
 jit_model_export_path=$3
 
 python3 -u ${BIN_DIR}/export.py \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --checkpoint_path ${ckpt_path_prefix} \
 --export_path ${jit_model_export_path}
diff --git a/examples/tiny/asr1/local/test.sh b/examples/tiny/asr1/local/test.sh
index 34088ce9..190bacff 100755
--- a/examples/tiny/asr1/local/test.sh
+++ b/examples/tiny/asr1/local/test.sh
@@ -31,7 +31,7 @@ for type in attention ctc_greedy_search; do
         batch_size=64
     fi
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${ckpt_prefix}.${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
@@ -48,7 +48,7 @@ for type in ctc_prefix_beam_search attention_rescoring; do
     echo "decoding ${type}"
     batch_size=1
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${ckpt_prefix}.${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/examples/tiny/asr1/local/train.sh b/examples/tiny/asr1/local/train.sh
index 71af3a00..1c8593bd 100755
--- a/examples/tiny/asr1/local/train.sh
+++ b/examples/tiny/asr1/local/train.sh
@@ -29,7 +29,7 @@ mkdir -p exp
 
 python3 -u ${BIN_DIR}/train.py \
 --seed ${seed} \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name} \
 --profiler-options "${profiler_options}" \
diff --git a/examples/wenetspeech/asr1/local/test.sh b/examples/wenetspeech/asr1/local/test.sh
index 47bd2f63..da159de7 100755
--- a/examples/wenetspeech/asr1/local/test.sh
+++ b/examples/wenetspeech/asr1/local/test.sh
@@ -34,7 +34,7 @@ for type in attention ctc_greedy_search; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
@@ -53,7 +53,7 @@ for type in ctc_prefix_beam_search attention_rescoring; do
     output_dir=${ckpt_prefix}
     mkdir -p ${output_dir}
     python3 -u ${BIN_DIR}/test.py \
-    --nproc ${ngpu} \
+    --ngpu ${ngpu} \
     --config ${config_path} \
     --result_file ${output_dir}/${type}.rsl \
     --checkpoint_path ${ckpt_prefix} \
diff --git a/paddlespeech/s2t/decoders/recog.py b/paddlespeech/s2t/decoders/recog.py
index d9324ca0..3e9939f0 100644
--- a/paddlespeech/s2t/decoders/recog.py
+++ b/paddlespeech/s2t/decoders/recog.py
@@ -40,7 +40,6 @@ def get_config(config_path):
 
 
 def load_trained_model(args):
-    args.nprocs = args.ngpu
     confs = get_config(args.model_conf)
     class_obj = dynamic_import_tester(args.model_name)
     exp = class_obj(confs, args)
diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/test_hub.py b/paddlespeech/s2t/exps/deepspeech2/bin/test_hub.py
index 8ab8fea2..831bd1ad 100644
--- a/paddlespeech/s2t/exps/deepspeech2/bin/test_hub.py
+++ b/paddlespeech/s2t/exps/deepspeech2/bin/test_hub.py
@@ -87,7 +87,7 @@ class DeepSpeech2Tester_hub():
     def setup(self):
         """Setup the experiment.
         """
-        paddle.set_device('gpu' if self.args.nprocs > 0 else 'cpu')
+        paddle.set_device('gpu' if self.args.ngpu > 0 else 'cpu')
 
         self.setup_output_dir()
         self.setup_checkpointer()
diff --git a/paddlespeech/s2t/exps/deepspeech2/bin/train.py b/paddlespeech/s2t/exps/deepspeech2/bin/train.py
index d9b610a0..400538f9 100644
--- a/paddlespeech/s2t/exps/deepspeech2/bin/train.py
+++ b/paddlespeech/s2t/exps/deepspeech2/bin/train.py
@@ -27,8 +27,8 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.nprocs > 0:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    if args.ngpu > 1:
+        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
 
diff --git a/paddlespeech/s2t/exps/u2/bin/test_wav.py b/paddlespeech/s2t/exps/u2/bin/test_wav.py
index e118b481..a9450129 100644
--- a/paddlespeech/s2t/exps/u2/bin/test_wav.py
+++ b/paddlespeech/s2t/exps/u2/bin/test_wav.py
@@ -47,7 +47,7 @@ class U2Infer():
             vocab_filepath=config.collator.vocab_filepath,
             spm_model_prefix=config.collator.spm_model_prefix)
 
-        paddle.set_device('gpu' if self.args.nprocs > 0 else 'cpu')
+        paddle.set_device('gpu' if self.args.ngpu > 0 else 'cpu')
 
         # model
         model_conf = config.model
diff --git a/paddlespeech/s2t/exps/u2/bin/train.py b/paddlespeech/s2t/exps/u2/bin/train.py
index 127db521..d6ee8b30 100644
--- a/paddlespeech/s2t/exps/u2/bin/train.py
+++ b/paddlespeech/s2t/exps/u2/bin/train.py
@@ -32,8 +32,8 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.nprocs > 0:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    if args.ngpu > 1:
+        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
 
diff --git a/paddlespeech/s2t/exps/u2/model.py b/paddlespeech/s2t/exps/u2/model.py
index 27bc47d2..b6dbcf44 100644
--- a/paddlespeech/s2t/exps/u2/model.py
+++ b/paddlespeech/s2t/exps/u2/model.py
@@ -257,7 +257,7 @@ class U2Trainer(Trainer):
                 maxlen_in=float('inf'),
                 maxlen_out=float('inf'),
                 minibatches=0,
-                mini_batch_size=self.args.nprocs,
+                mini_batch_size=self.args.ngpu,
                 batch_count='auto',
                 batch_bins=0,
                 batch_frames_in=0,
@@ -277,7 +277,7 @@ class U2Trainer(Trainer):
                 maxlen_in=float('inf'),
                 maxlen_out=float('inf'),
                 minibatches=0,
-                mini_batch_size=self.args.nprocs,
+                mini_batch_size=self.args.ngpu,
                 batch_count='auto',
                 batch_bins=0,
                 batch_frames_in=0,
diff --git a/paddlespeech/s2t/exps/u2_kaldi/bin/train.py b/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
index d3427eec..fcfc05a8 100644
--- a/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
+++ b/paddlespeech/s2t/exps/u2_kaldi/bin/train.py
@@ -36,8 +36,8 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.nprocs > 0:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    if args.ngpu > 1:
+        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
 
diff --git a/paddlespeech/s2t/exps/u2_kaldi/model.py b/paddlespeech/s2t/exps/u2_kaldi/model.py
index d82034c8..c23b4c24 100644
--- a/paddlespeech/s2t/exps/u2_kaldi/model.py
+++ b/paddlespeech/s2t/exps/u2_kaldi/model.py
@@ -239,7 +239,7 @@ class U2Trainer(Trainer):
             maxlen_in=float('inf'),
             maxlen_out=float('inf'),
             minibatches=0,
-            mini_batch_size=self.args.nprocs,
+            mini_batch_size=self.args.ngpu,
             batch_count='auto',
             batch_bins=0,
             batch_frames_in=0,
@@ -258,7 +258,7 @@ class U2Trainer(Trainer):
             maxlen_in=float('inf'),
             maxlen_out=float('inf'),
             minibatches=0,
-            mini_batch_size=self.args.nprocs,
+            mini_batch_size=self.args.ngpu,
             batch_count='auto',
             batch_bins=0,
             batch_frames_in=0,
diff --git a/paddlespeech/s2t/exps/u2_st/bin/train.py b/paddlespeech/s2t/exps/u2_st/bin/train.py
index 3d823cc4..58496c88 100644
--- a/paddlespeech/s2t/exps/u2_st/bin/train.py
+++ b/paddlespeech/s2t/exps/u2_st/bin/train.py
@@ -30,8 +30,8 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.nprocs > 0:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    if args.ngpu > 1:
+        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
 
diff --git a/paddlespeech/s2t/training/cli.py b/paddlespeech/s2t/training/cli.py
index 55b010e9..3ef871c5 100644
--- a/paddlespeech/s2t/training/cli.py
+++ b/paddlespeech/s2t/training/cli.py
@@ -51,7 +51,7 @@ def default_argument_parser(parser=None):
 
     The ``--checkpoint_path`` specifies the checkpoint to load from.
 
-    The ``--nprocs`` specifies how to run the training.
+    The ``--ngpu`` specifies how to run the training.
 
 
     See Also
@@ -78,7 +78,7 @@ def default_argument_parser(parser=None):
         help="seed to use for paddle, np and random. None or 0 for random, else set seed."
     )
     train_group.add_argument(
-        "--nprocs",
+        "--ngpu",
         type=int,
         default=1,
         help="number of parallel processes. 0 for cpu.")
diff --git a/paddlespeech/s2t/training/trainer.py b/paddlespeech/s2t/training/trainer.py
index e6328cdf..f5fb2db0 100644
--- a/paddlespeech/s2t/training/trainer.py
+++ b/paddlespeech/s2t/training/trainer.py
@@ -88,8 +88,8 @@ class Trainer():
     >>>     config.merge_from_list(args.opts)
     >>> config.freeze()
     >>>
-    >>> if args.nprocs > 0:
-    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    >>> if args.ngpu > 1:
+    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     >>> else:
     >>>     main_sp(config, args)
     """
@@ -112,7 +112,7 @@ class Trainer():
         logger.info(f"Rank: {self.rank}/{self.world_size}")
 
         # set device
-        paddle.set_device('gpu' if self.args.nprocs > 0 else 'cpu')
+        paddle.set_device('gpu' if self.args.ngpu > 0 else 'cpu')
         if self.parallel:
             self.init_parallel()
 
@@ -162,7 +162,7 @@ class Trainer():
         """A flag indicating whether the experiment should run with
         multiprocessing.
         """
-        return self.args.nprocs > 1
+        return self.args.ngpu > 1
 
     def init_parallel(self):
         """Init environment for multiprocess training.
diff --git a/paddlespeech/t2s/exps/voice_cloning/tacotron2_ge2e/train.py b/paddlespeech/t2s/exps/voice_cloning/tacotron2_ge2e/train.py
index 8af1d45e..ea5f12da 100644
--- a/paddlespeech/t2s/exps/voice_cloning/tacotron2_ge2e/train.py
+++ b/paddlespeech/t2s/exps/voice_cloning/tacotron2_ge2e/train.py
@@ -241,7 +241,7 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.ngpu:
+    if args.ngpu > 1:
         dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
diff --git a/paddlespeech/t2s/training/cli.py b/paddlespeech/t2s/training/cli.py
index a0710fd7..83dae117 100644
--- a/paddlespeech/t2s/training/cli.py
+++ b/paddlespeech/t2s/training/cli.py
@@ -30,7 +30,7 @@ def default_argument_parser():
     
     The ``--checkpoint_path`` specifies the checkpoint to load from.
     
-    The ``--device`` and ``--nprocs`` specifies how to run the training.
+    The ``--ngpu`` specifies how to run the training.
     
     See Also
     --------
diff --git a/paddlespeech/t2s/training/experiment.py b/paddlespeech/t2s/training/experiment.py
index c9e7f4cc..de36db24 100644
--- a/paddlespeech/t2s/training/experiment.py
+++ b/paddlespeech/t2s/training/experiment.py
@@ -82,8 +82,8 @@ class ExperimentBase(object):
     >>>     config.merge_from_list(args.opts)
     >>> config.freeze()
     >>>
-    >>> if args.nprocs > 1 and args.device == "gpu":
-    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    >>> if args.ngpu > 1:
+    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     >>> else:
     >>>     main_sp(config, args)
     """
diff --git a/paddlespeech/text/examples/punctuation_restoration/chinese/local/test.sh b/paddlespeech/text/examples/punctuation_restoration/chinese/local/test.sh
index 6db75ca2..ee022462 100644
--- a/paddlespeech/text/examples/punctuation_restoration/chinese/local/test.sh
+++ b/paddlespeech/text/examples/punctuation_restoration/chinese/local/test.sh
@@ -9,17 +9,11 @@ fi
 ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
 echo "using $ngpu gpus..."
 
-device=gpu
-if [ ${ngpu} == 0 ];then
-    device=cpu
-fi
 config_path=$1
 ckpt_prefix=$2
 
-
 python3 -u ${BIN_DIR}/test.py \
---device ${device} \
---nproc 1 \
+--ngpu 1 \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix}
diff --git a/paddlespeech/text/examples/punctuation_restoration/chinese/local/train.sh b/paddlespeech/text/examples/punctuation_restoration/chinese/local/train.sh
index f6bd2c98..fc345cc1 100644
--- a/paddlespeech/text/examples/punctuation_restoration/chinese/local/train.sh
+++ b/paddlespeech/text/examples/punctuation_restoration/chinese/local/train.sh
@@ -11,16 +11,10 @@ echo "using $ngpu gpus..."
 config_path=$1
 ckpt_name=$2
 
-device=gpu
-if [ ${ngpu} == 0 ];then
-    device=cpu
-fi
-
 mkdir -p exp
 
 python3 -u ${BIN_DIR}/train.py \
---device ${device} \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name}
 
diff --git a/paddlespeech/text/examples/punctuation_restoration/english/local/test.sh b/paddlespeech/text/examples/punctuation_restoration/english/local/test.sh
index 6db75ca2..d8a58f34 100644
--- a/paddlespeech/text/examples/punctuation_restoration/english/local/test.sh
+++ b/paddlespeech/text/examples/punctuation_restoration/english/local/test.sh
@@ -9,17 +9,12 @@ fi
 ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
 echo "using $ngpu gpus..."
 
-device=gpu
-if [ ${ngpu} == 0 ];then
-    device=cpu
-fi
 config_path=$1
 ckpt_prefix=$2
 
 
 python3 -u ${BIN_DIR}/test.py \
---device ${device} \
---nproc 1 \
+--ngpu 1 \
 --config ${config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix}
diff --git a/paddlespeech/text/examples/punctuation_restoration/english/local/train.sh b/paddlespeech/text/examples/punctuation_restoration/english/local/train.sh
index f6bd2c98..fc345cc1 100644
--- a/paddlespeech/text/examples/punctuation_restoration/english/local/train.sh
+++ b/paddlespeech/text/examples/punctuation_restoration/english/local/train.sh
@@ -11,16 +11,10 @@ echo "using $ngpu gpus..."
 config_path=$1
 ckpt_name=$2
 
-device=gpu
-if [ ${ngpu} == 0 ];then
-    device=cpu
-fi
-
 mkdir -p exp
 
 python3 -u ${BIN_DIR}/train.py \
---device ${device} \
---nproc ${ngpu} \
+--ngpu ${ngpu} \
 --config ${config_path} \
 --output exp/${ckpt_name}
 
diff --git a/paddlespeech/text/speechtask/punctuation_restoration/bin/train.py b/paddlespeech/text/speechtask/punctuation_restoration/bin/train.py
index 1ffd79b7..c4b67265 100644
--- a/paddlespeech/text/speechtask/punctuation_restoration/bin/train.py
+++ b/paddlespeech/text/speechtask/punctuation_restoration/bin/train.py
@@ -26,8 +26,8 @@ def main_sp(config, args):
 
 
 def main(config, args):
-    if args.device == "gpu" and args.nprocs > 1:
-        dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    if args.ngpu > 1:
+        dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     else:
         main_sp(config, args)
 
diff --git a/paddlespeech/text/speechtask/punctuation_restoration/training/trainer.py b/paddlespeech/text/speechtask/punctuation_restoration/training/trainer.py
index 2dce88a3..d6b6eeb6 100644
--- a/paddlespeech/text/speechtask/punctuation_restoration/training/trainer.py
+++ b/paddlespeech/text/speechtask/punctuation_restoration/training/trainer.py
@@ -106,8 +106,8 @@ class Trainer():
     >>>     config.merge_from_list(args.opts)
     >>> config.freeze()
     >>> 
-    >>> if args.nprocs > 1 and args.device == "gpu":
-    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.nprocs)
+    >>> if args.ngpu > 1:
+    >>>     dist.spawn(main_sp, args=(config, args), nprocs=args.ngpu)
     >>> else:
     >>>     main_sp(config, args)
     """
@@ -147,7 +147,7 @@ class Trainer():
         """A flag indicating whether the experiment should run with 
         multiprocessing.
         """
-        return self.args.device == "gpu" and self.args.nprocs > 1
+        return self.args.ngpu > 1
 
     def init_parallel(self):
         """Init environment for multiprocess training.
diff --git a/paddlespeech/text/speechtask/punctuation_restoration/utils/default_parser.py b/paddlespeech/text/speechtask/punctuation_restoration/utils/default_parser.py
index 405b29a2..de0d8aef 100644
--- a/paddlespeech/text/speechtask/punctuation_restoration/utils/default_parser.py
+++ b/paddlespeech/text/speechtask/punctuation_restoration/utils/default_parser.py
@@ -30,7 +30,7 @@ def default_argument_parser():
     
     The ``--checkpoint_path`` specifies the checkpoint to load from.
     
-    The ``--device`` and ``--nprocs`` specifies how to run the training.
+    The ``--ngpu`` specifies how to run the training.
     
     
     See Also
@@ -60,9 +60,7 @@ def default_argument_parser():
     parser.add_argument("--result_file", type=str, help="path of save the asr result")
 
     # running
-    parser.add_argument("--device", type=str, default='gpu', choices=["cpu", "gpu"],
-                        help="device type to use, cpu and gpu are supported.")
-    parser.add_argument("--nprocs", type=int, default=1, help="number of parallel processes to use.")
+    parser.add_argument("--ngpu", type=int, default=1, help="number of parallel processes to use. if ngpu=0, using cpu.")
 
     # overwrite extra config and default config
     # parser.add_argument("--opts", nargs=argparse.REMAINDER, 
diff --git a/tests/benchmark/conformer/run_benchmark.sh b/tests/benchmark/conformer/run_benchmark.sh
index 56b63e76..d4efe2b9 100644
--- a/tests/benchmark/conformer/run_benchmark.sh
+++ b/tests/benchmark/conformer/run_benchmark.sh
@@ -38,7 +38,7 @@ function _train(){
     train_cmd="--config=${config_path}
                --output=${output}
                --seed=${seed}
-               --nproc=${ngpu}
+               --ngpu=${ngpu}
                --profiler-options "${profiler_options}"
                --benchmark-batch-size ${batch_size}
                --benchmark-max-step ${benchmark_max_step} "
diff --git a/tests/chains/ds2/ds2_params_lite_train_infer.txt b/tests/chains/ds2/ds2_params_lite_train_infer.txt
index c1cbfbb9..b11872bd 100644
--- a/tests/chains/ds2/ds2_params_lite_train_infer.txt
+++ b/tests/chains/ds2/ds2_params_lite_train_infer.txt
@@ -21,13 +21,13 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval: ../../../paddlespeech/s2t/exps/deepspeech2/bin/test.py --nproc 1 --config conf/deepspeech2.yaml --checkpoint_path exp/deepspeech_tiny/checkpoints/9 --result_file tests/9.rsl  --model_type offline
+eval: ../../../paddlespeech/s2t/exps/deepspeech2/bin/test.py --ngpu 1 --config conf/deepspeech2.yaml --checkpoint_path exp/deepspeech_tiny/checkpoints/9 --result_file tests/9.rsl  --model_type offline
 null:null
 ##
 ===========================infer_params===========================
 null:null
 null:null
-norm_export: ../../../paddlespeech/s2t/exps/deepspeech2/bin/export.py --nproc 1 --config conf/deepspeech2.yaml --model_type offline --checkpoint_path exp/deepspeech_tiny/checkpoints/9 --export_path exp/deepspeech_tiny/checkpoints/9.jit
+norm_export: ../../../paddlespeech/s2t/exps/deepspeech2/bin/export.py --ngpu 1 --config conf/deepspeech2.yaml --model_type offline --checkpoint_path exp/deepspeech_tiny/checkpoints/9 --export_path exp/deepspeech_tiny/checkpoints/9.jit
 quant_export:null
 fpgm_export:null
 distill_export:null
diff --git a/tests/chains/ds2/ds2_params_whole_train_infer.txt b/tests/chains/ds2/ds2_params_whole_train_infer.txt
index bfcb745f..875e3ccf 100644
--- a/tests/chains/ds2/ds2_params_whole_train_infer.txt
+++ b/tests/chains/ds2/ds2_params_whole_train_infer.txt
@@ -21,13 +21,13 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval: ../../../paddlespeech/s2t/exps/deepspeech2/bin/test.py --nproc 1 --config conf/deepspeech2.yaml --result_file tests/49.rsl --checkpoint_path exp/deepspeech_whole/checkpoints/49 --model_type offline
+eval: ../../../paddlespeech/s2t/exps/deepspeech2/bin/test.py --ngpu 1 --config conf/deepspeech2.yaml --result_file tests/49.rsl --checkpoint_path exp/deepspeech_whole/checkpoints/49 --model_type offline
 null:null
 ##
 ===========================infer_params===========================
 null:null
 null:null
-norm_export: ../../../paddlespeech/s2t/exps/deepspeech2/bin/export.py --nproc 1 --config conf/deepspeech2.yaml --model_type offline  --checkpoint_path exp/deepspeech_whole/checkpoints/49 --export_path exp/deepspeech_whole/checkpoints/49.jit
+norm_export: ../../../paddlespeech/s2t/exps/deepspeech2/bin/export.py --ngpu 1 --config conf/deepspeech2.yaml --model_type offline  --checkpoint_path exp/deepspeech_whole/checkpoints/49 --export_path exp/deepspeech_whole/checkpoints/49.jit
 quant_export:null
 fpgm_export:null
 distill_export:null
diff --git a/tests/chains/ds2/speedyspeech_params_lite.txt b/tests/chains/ds2/speedyspeech_params_lite.txt
index c1cfb8f5..487b0b5e 100644
--- a/tests/chains/ds2/speedyspeech_params_lite.txt
+++ b/tests/chains/ds2/speedyspeech_params_lite.txt
@@ -21,7 +21,7 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:../examples/speedyspeech/baker/synthesize_e2e.py --speedyspeech-config=../examples/speedyspeech/baker/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_90.pdz --speedyspeech-stat=pretrain_models/speedyspeech_baker_ckpt_0.4/speedy_speech_stats.npy --pwg-config=../examples/parallelwave_gan/baker/conf/default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../examples/speedyspeech/baker/sentences.txt --output-dir=e2e --inference-dir=inference --device="gpu" --phones-dict=../examples/speedyspeech/baker/phones.txt --tones-dict=../examples/speedyspeech/baker/tones.txt
+eval:../examples/speedyspeech/baker/synthesize_e2e.py --speedyspeech-config=../examples/speedyspeech/baker/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_90.pdz --speedyspeech-stat=pretrain_models/speedyspeech_baker_ckpt_0.4/speedy_speech_stats.npy --pwg-config=../examples/parallelwave_gan/baker/conf/default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../examples/speedyspeech/baker/sentences.txt --output-dir=e2e --inference-dir=inference --ngpu=1 --phones-dict=../examples/speedyspeech/baker/phones.txt --tones-dict=../examples/speedyspeech/baker/tones.txt
 null:null
 ##
 ===========================infer_params===========================
diff --git a/tests/chains/ds2/test.sh b/tests/chains/ds2/test.sh
index 0b2b4f58..c9307820 100644
--- a/tests/chains/ds2/test.sh
+++ b/tests/chains/ds2/test.sh
@@ -323,7 +323,7 @@ else
                 elif [ ${#gpu} -le 15 ];then  # train with multi-gpu
                     gsu=${gpu//,/ }
                     nump=`echo $gsu | wc -w`
-                    cmd="${python} ${run_train} --nproc=$nump"
+                    cmd="${python} ${run_train} --ngpu=$nump"
                 else     # train with multi-machine
                     cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1}"
                 fi
diff --git a/tests/chains/speedyspeech/speedyspeech_params_lite_multi_gpu.txt b/tests/chains/speedyspeech/speedyspeech_params_lite_multi_gpu.txt
index 0f64da27..634c3a5a 100644
--- a/tests/chains/speedyspeech/speedyspeech_params_lite_multi_gpu.txt
+++ b/tests/chains/speedyspeech/speedyspeech_params_lite_multi_gpu.txt
@@ -21,7 +21,7 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=../../../examples/csmsc/tts2/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_20.pdz --speedyspeech-stat=train_data/mini_BZNSYP/train/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --device="gpu" --phones-dict=train_data/mini_BZNSYP/phone_id_map.txt --tones-dict=train_data/mini_BZNSYP/tone_id_map.txt
+eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=../../../examples/csmsc/tts2/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_20.pdz --speedyspeech-stat=train_data/mini_BZNSYP/train/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --ngpu=1 --phones-dict=train_data/mini_BZNSYP/phone_id_map.txt --tones-dict=train_data/mini_BZNSYP/tone_id_map.txt
 null:null
 ##
 ===========================infer_params===========================
diff --git a/tests/chains/speedyspeech/speedyspeech_params_lite_single_gpu.txt b/tests/chains/speedyspeech/speedyspeech_params_lite_single_gpu.txt
index beda4c04..d187d4c6 100644
--- a/tests/chains/speedyspeech/speedyspeech_params_lite_single_gpu.txt
+++ b/tests/chains/speedyspeech/speedyspeech_params_lite_single_gpu.txt
@@ -21,7 +21,7 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=../../../examples/csmsc/tts2/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_30.pdz --speedyspeech-stat=train_data/mini_BZNSYP/train/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --device="gpu" --phones-dict=train_data/mini_BZNSYP/phone_id_map.txt --tones-dict=train_data/mini_BZNSYP/tone_id_map.txt
+eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=../../../examples/csmsc/tts2/conf/default.yaml --speedyspeech-checkpoint=exp/default/checkpoints/snapshot_iter_30.pdz --speedyspeech-stat=train_data/mini_BZNSYP/train/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --ngpu=1 --phones-dict=train_data/mini_BZNSYP/phone_id_map.txt --tones-dict=train_data/mini_BZNSYP/tone_id_map.txt
 null:null
 ##
 ===========================infer_params===========================
diff --git a/tests/chains/speedyspeech/speedyspeech_params_whole_multi_gpu.txt b/tests/chains/speedyspeech/speedyspeech_params_whole_multi_gpu.txt
index ecdbf76d..f8ebd499 100644
--- a/tests/chains/speedyspeech/speedyspeech_params_whole_multi_gpu.txt
+++ b/tests/chains/speedyspeech/speedyspeech_params_whole_multi_gpu.txt
@@ -21,7 +21,7 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/default.yaml --speedyspeech-checkpoint=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/snapshot_iter_11400.pdz --speedyspeech-stat=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --device="gpu" --phones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/phone_id_map.txt --tones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/tone_id_map.txt
+eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/default.yaml --speedyspeech-checkpoint=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/snapshot_iter_11400.pdz --speedyspeech-stat=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --ngpu=1 --phones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/phone_id_map.txt --tones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/tone_id_map.txt
 null:null
 ##
 ===========================infer_params===========================
diff --git a/tests/chains/speedyspeech/speedyspeech_params_whole_single_gpu.txt b/tests/chains/speedyspeech/speedyspeech_params_whole_single_gpu.txt
index 523b5c6e..0e3c49cb 100644
--- a/tests/chains/speedyspeech/speedyspeech_params_whole_single_gpu.txt
+++ b/tests/chains/speedyspeech/speedyspeech_params_whole_single_gpu.txt
@@ -21,7 +21,7 @@ null:null
 null:null
 ##
 ===========================eval_params===========================
-eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/default.yaml --speedyspeech-checkpoint=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/snapshot_iter_11400.pdz --speedyspeech-stat=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --device="gpu" --phones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/phone_id_map.txt --tones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/tone_id_map.txt
+eval:../../../paddlespeech/t2s/exps/speedyspeech/synthesize_e2e.py --speedyspeech-config=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/default.yaml --speedyspeech-checkpoint=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/snapshot_iter_11400.pdz --speedyspeech-stat=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/feats_stats.npy --pwg-config=pretrain_models/pwg_baker_ckpt_0.4/pwg_default.yaml --pwg-checkpoint=pretrain_models/pwg_baker_ckpt_0.4/pwg_snapshot_iter_400000.pdz --pwg-stat=pretrain_models/pwg_baker_ckpt_0.4/pwg_stats.npy --text=../../../paddlespeech/t2s/exps/sentences.txt --output-dir=e2e --inference-dir=inference --ngpu=1 --phones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/phone_id_map.txt --tones-dict=pretrain_models/speedyspeech_nosil_baker_ckpt_0.5/tone_id_map.txt
 null:null
 ##
 ===========================infer_params===========================