merge dygraph

835e7178 · WenmuZhou · 55f6a634 · 63ed5fca · 835e7178 · 835e7178
Commit 835e7178 authored Aug 31, 2021 by WenmuZhou
12 changed files
--- a/ppstructure/table/README_ch.md
+++ b/ppstructure/table/README_ch.md
@@ -43,7 +43,7 @@ wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_tab
 wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_structure_infer.tar && tar xf en_ppocr_mobile_v2.0_table_structure_infer.tar
 cd ..
 # 执行预测
-python3 table/predict_table.py --det_model_dir=inference/en_ppocr_mobile_v2.0_table_det_infer --rec_model_dir=inference/en_ppocr_mobile_v2.0_table_rec_infer --table_model_dir=inference/en_ppocr_mobile_v2.0_table_structure_infer --image_dir=../doc/table/table.jpg --rec_char_dict_path=../ppocr/utils/ppocr_keys_v1.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=ch --det_limit_side_len=736 --det_limit_type=min --output ../output/table
+python3 table/predict_table.py --det_model_dir=inference/en_ppocr_mobile_v2.0_table_det_infer --rec_model_dir=inference/en_ppocr_mobile_v2.0_table_rec_infer --table_model_dir=inference/en_ppocr_mobile_v2.0_table_structure_infer --image_dir=../doc/table/table.jpg --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=EN --det_limit_side_len=736 --det_limit_type=min --output ../output/table
 ```
 运行完成后，每张图片的excel表格会保存到output字段指定的目录下

--- a/requirements.txt
+++ b/requirements.txt
@@ -9,3 +9,6 @@ visualdl
 python-Levenshtein
 opencv-contrib-python==4.4.0.46
 cython
+lxml
+premailer
+openpyxl
\ No newline at end of file
--- a/tests/ocr_det_params.txt
+++ b/tests/ocr_det_params.txt
@@ -4,7 +4,7 @@ python:python3.7
 gpu_list:0|0,1
 Global.use_gpu:True|True
 Global.auto_cast:null
-Global.epoch_num:lite_train_infer=2|whole_train_infer=300
+Global.epoch_num:lite_train_infer=1|whole_train_infer=300
 Global.save_model_dir:./output/
 Train.loader.batch_size_per_card:lite_train_infer=2|whole_train_infer=4
 Global.pretrained_model:null
@@ -15,7 +15,7 @@ null:null
 trainer:norm_train|pact_train
 norm_train:tools/train.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained
 pact_train:deploy/slim/quantization/quant.py -c configs/det/det_mv3_db.yml -o
-fpgm_train:null
+fpgm_train:deploy/slim/prune/sensitivity_anal.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/det_mv3_db_v2.0_train/best_accuracy
 distill_train:null
 null:null
 null:null
@@ -29,7 +29,7 @@ Global.save_inference_dir:./output/
 Global.pretrained_model:
 norm_export:tools/export_model.py -c configs/det/det_mv3_db.yml -o 
 quant_export:deploy/slim/quantization/export_model.py -c configs/det/det_mv3_db.yml -o 
-fpgm_export:deploy/slim/prune/export_prune_model.py
+fpgm_export:deploy/slim/prune/export_prune_model.py -c configs/det/det_mv3_db.yml -o 
 distill_export:null
 export1:null
 export2:null
@@ -49,4 +49,19 @@ inference:tools/infer/predict_det.py
 --save_log_path:null
 --benchmark:True
 null:null
+===========================cpp_infer_params===========================
+use_opencv:True
+infer_model:./inference/ch_ppocr_mobile_v2.0_det_infer/
+infer_quant:False
+inference:./deploy/cpp_infer/build/ppocr det
+--use_gpu:True|False
+--enable_mkldnn:True|False
+--cpu_threads:1|6
+--rec_batch_num:1
+--use_tensorrt:False|True
+--precision:fp32|fp16
+--det_model_dir:
+--image_dir:./inference/ch_det_data_50/all-sum-510/
+--save_log_path:null
+--benchmark:True
--- a/tests/ocr_det_server_params.txt
+++ b/tests/ocr_det_server_params.txt
+===========================train_params===========================
+model_name:ocr_server_det
+python:python3.7
+gpu_list:0|0,1
+Global.use_gpu:True|True
+Global.auto_cast:null
+Global.epoch_num:lite_train_infer=2|whole_train_infer=300
+Global.save_model_dir:./output/
+Train.loader.batch_size_per_card:lite_train_infer=2|whole_train_infer=4
+Global.pretrained_model:null
+train_model_name:latest
+train_infer_img_dir:./train_data/icdar2015/text_localization/ch4_test_images/
+null:null
+##
+trainer:norm_train|pact_train
+norm_train:tools/train.py -c configs/det/det_r50_vd_db.yml -o Global.pretrained_model=""
+pact_train:null
+fpgm_train:null
+distill_train:null
+null:null
+null:null
+##
+===========================eval_params=========================== 
+eval:tools/eval.py -c configs/det/det_mv3_db.yml -o
+null:null
+##
+===========================infer_params===========================
+Global.save_inference_dir:./output/
+Global.pretrained_model:
+norm_export:tools/export_model.py -c configs/det/det_r50_vd_db.yml -o 
+quant_export:null 
+fpgm_export:null
+distill_export:null
+export1:null
+export2:null
+##
+infer_model:./inference/ch_ppocr_server_v2.0_det_infer/
+infer_export:null
+infer_quant:False
+inference:tools/infer/predict_det.py
+--use_gpu:True|False
+--enable_mkldnn:True|False
+--cpu_threads:1|6
+--rec_batch_num:1
+--use_tensorrt:False|True
+--precision:fp32|fp16|int8
+--det_model_dir:
+--image_dir:./inference/ch_det_data_50/all-sum-510/
+--save_log_path:null
+--benchmark:True
+null:null
--- a/tests/prepare.sh
+++ b/tests/prepare.sh
 #!/bin/bash
 FILENAME=$1
-# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer']
+# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'cpp_infer']
 MODE=$2
 dataline=$(cat ${FILENAME})
@@ -34,10 +34,13 @@ MODE=$2
 if [ ${MODE} = "lite_train_infer" ];then
    # pretrain lite train data
    wget -nc -P  ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams
+    wget -nc -P ./pretrain_models/  https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/det_mv3_db_v2.0_train.tar
+    cd ./pretrain_models/ && tar xf det_mv3_db_v2.0_train.tar && cd ../
    rm -rf ./train_data/icdar2015
    rm -rf ./train_data/ic15_data
    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_lite.tar
    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar # todo change to bcebos
+    wget -nc -P ./deploy/slim/prune https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/sen.pickle
    cd ./train_data/ && tar xf icdar2015_lite.tar && tar xf ic15_data.tar
    ln -s ./icdar2015_lite ./icdar2015
@@ -58,13 +61,17 @@ elif [ ${MODE} = "whole_infer" ];then
    cd ./train_data/ && tar xf icdar2015_infer.tar && tar xf ic15_data.tar
    ln -s ./icdar2015_infer ./icdar2015
    cd ../
-else
+elif [ ${MODE} = "infer" ] || [ ${MODE} = "cpp_infer" ];then
    if [ ${model_name} = "ocr_det" ]; then
        eval_model_name="ch_ppocr_mobile_v2.0_det_infer"
        rm -rf ./train_data/icdar2015
-        wget -nc -P ./train_data https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
+        wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
        wget -nc  -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
        cd ./inference && tar xf ${eval_model_name}.tar && tar xf ch_det_data_50.tar && cd ../
+    elif [ ${model_name} = "ocr_server_det" ]; then
+        wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_server_v2.0_det_infer.tar
+        wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
+        cd ./inference && tar xf ch_ppocr_server_v2.0_det_infer.tar && tar xf ch_det_data_50.tar && cd ../
    else 
        rm -rf ./train_data/ic15_data
        eval_model_name="ch_ppocr_mobile_v2.0_rec_infer"
@@ -74,3 +81,72 @@ else
    fi 
 fi
+if [ ${MODE} = "cpp_infer" ];then
+    cd deploy/cpp_infer
+    use_opencv=$(func_parser_value "${lines[52]}")
+    if [ ${use_opencv} = "True" ]; then
+        echo "################### build opencv ###################"
+        rm -rf 3.4.7.tar.gz opencv-3.4.7/
+        wget https://github.com/opencv/opencv/archive/3.4.7.tar.gz
+        tar -xf 3.4.7.tar.gz
+        cd opencv-3.4.7/
+        install_path=$(pwd)/opencv-3.4.7/opencv3
+        rm -rf build
+        mkdir build
+        cd build
+        cmake .. \
+            -DCMAKE_INSTALL_PREFIX=${install_path} \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DBUILD_SHARED_LIBS=OFF \
+            -DWITH_IPP=OFF \
+            -DBUILD_IPP_IW=OFF \
+            -DWITH_LAPACK=OFF \
+            -DWITH_EIGEN=OFF \
+            -DCMAKE_INSTALL_LIBDIR=lib64 \
+            -DWITH_ZLIB=ON \
+            -DBUILD_ZLIB=ON \
+            -DWITH_JPEG=ON \
+            -DBUILD_JPEG=ON \
+            -DWITH_PNG=ON \
+            -DBUILD_PNG=ON \
+            -DWITH_TIFF=ON \
+            -DBUILD_TIFF=ON
+        make -j
+        make install
+        cd ../
+        echo "################### build opencv finished ###################"
+    fi
+    echo "################### build PaddleOCR demo ####################"
+    if [ ${use_opencv} = "True" ]; then
+        OPENCV_DIR=$(pwd)/opencv-3.4.7/opencv3/
+    else
+        OPENCV_DIR=''
+    fi
+    LIB_DIR=$(pwd)/Paddle/build/paddle_inference_install_dir/
+    CUDA_LIB_DIR=$(dirname `find /usr -name libcudart.so`)
+    CUDNN_LIB_DIR=$(dirname `find /usr -name libcudnn.so`)
+    BUILD_DIR=build
+    rm -rf ${BUILD_DIR}
+    mkdir ${BUILD_DIR}
+    cd ${BUILD_DIR}
+    cmake .. \
+        -DPADDLE_LIB=${LIB_DIR} \
+        -DWITH_MKL=ON \
+        -DWITH_GPU=OFF \
+        -DWITH_STATIC_LIB=OFF \
+        -DWITH_TENSORRT=OFF \
+        -DOPENCV_DIR=${OPENCV_DIR} \
+        -DCUDNN_LIB=${CUDNN_LIB_DIR} \
+        -DCUDA_LIB=${CUDA_LIB_DIR} \
+        -DTENSORRT_DIR=${TENSORRT_DIR} \
+    make -j
+    echo "################### build PaddleOCR demo finished ###################"
+fi
\ No newline at end of file
--- a/tests/readme.md
+++ b/tests/readme.md
+# 介绍
+test.sh和params.txt文件配合使用，完成OCR轻量检测和识别模型从训练到预测的流程测试。
+# 安装依赖
+- 安装PaddlePaddle >= 2.0
+- 安装PaddleOCR依赖
+    ```
+    pip3 install  -r ../requirements.txt
+    ```
+- 安装autolog
+    ```
+    git clone https://github.com/LDOUBLEV/AutoLog
+    cd AutoLog
+    pip3 install -r requirements.txt
+    python3 setup.py bdist_wheel
+    pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl
+    cd ../
+    ```
+# 目录介绍
+```bash
+tests/
+├── ocr_det_params.txt   # 测试OCR检测模型的参数配置文件
+├── ocr_rec_params.txt   # 测试OCR识别模型的参数配置文件
+└── prepare.sh           # 完成test.sh运行所需要的数据和模型下载
+└── test.sh              # 根据
+```
+# 使用方法
+test.sh包含四种运行模式，每种模式的运行数据不同，分别用于测试速度和精度，分别是：
+- 模式1 lite_train_infer，使用少量数据训练，用于快速验证训练到预测的走通流程，不验证精度和速度；
+```
+bash test/prepare.sh ./tests/ocr_det_params.txt 'lite_train_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'lite_train_infer'
+```
+- 模式2 whole_infer，使用少量数据训练，一定量数据预测，用于验证训练后的模型执行预测，预测速度是否合理；
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'whole_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'whole_infer'
+```
+- 模式3 infer 不训练，全量数据预测，走通开源模型评估、动转静，检查inference model预测时间和精度;
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'infer'
+用法1:
+bash tests/test.sh ./tests/ocr_det_params.txt 'infer'
+用法2: 指定GPU卡预测，第三个传入参数为GPU卡号
+bash tests/test.sh ./tests/ocr_det_params.txt 'infer' '1'
+```
+模式4: whole_train_infer , CE： 全量数据训练，全量数据预测，验证模型训练精度，预测精度，预测速度
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'whole_train_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'whole_train_infer'
+```
--- a/tests/test.sh
+++ b/tests/test.sh
 #!/bin/bash
 FILENAME=$1
-# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer']
+# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'cpp_infer']
 MODE=$2
 dataline=$(cat ${FILENAME})
@@ -145,6 +145,33 @@ benchmark_value=$(func_parser_value "${lines[49]}")
 infer_key1=$(func_parser_key "${lines[50]}")
 infer_value1=$(func_parser_value "${lines[50]}")
+if [ ${MODE} = "cpp_infer" ]; then
+    # parser cpp inference model 
+    cpp_infer_model_dir_list=$(func_parser_value "${lines[53]}")
+    cpp_infer_is_quant=$(func_parser_value "${lines[54]}")
+    # parser cpp inference 
+    inference_cmd=$(func_parser_value "${lines[55]}")
+    cpp_use_gpu_key=$(func_parser_key "${lines[56]}")
+    cpp_use_gpu_list=$(func_parser_value "${lines[56]}")
+    cpp_use_mkldnn_key=$(func_parser_key "${lines[57]}")
+    cpp_use_mkldnn_list=$(func_parser_value "${lines[57]}")
+    cpp_cpu_threads_key=$(func_parser_key "${lines[58]}")
+    cpp_cpu_threads_list=$(func_parser_value "${lines[58]}")
+    cpp_batch_size_key=$(func_parser_key "${lines[59]}")
+    cpp_batch_size_list=$(func_parser_value "${lines[59]}")
+    cpp_use_trt_key=$(func_parser_key "${lines[60]}")
+    cpp_use_trt_list=$(func_parser_value "${lines[60]}")
+    cpp_precision_key=$(func_parser_key "${lines[61]}")
+    cpp_precision_list=$(func_parser_value "${lines[61]}")
+    cpp_infer_model_key=$(func_parser_key "${lines[62]}")
+    cpp_image_dir_key=$(func_parser_key "${lines[63]}")
+    cpp_infer_img_dir=$(func_parser_value "${lines[63]}")
+    cpp_save_log_key=$(func_parser_key "${lines[64]}")
+    cpp_benchmark_key=$(func_parser_key "${lines[65]}")
+    cpp_benchmark_value=$(func_parser_value "${lines[65]}")
+fi
 LOG_PATH="./tests/output"
 mkdir -p ${LOG_PATH}
 status_log="${LOG_PATH}/results.log"
@@ -218,6 +245,71 @@ function func_inference(){
    done
 }
+function func_cpp_inference(){
+    IFS='|'
+    _script=$1
+    _model_dir=$2
+    _log_path=$3
+    _img_dir=$4
+    _flag_quant=$5
+    # inference 
+    for use_gpu in ${cpp_use_gpu_list[*]}; do
+        if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then
+            for use_mkldnn in ${cpp_use_mkldnn_list[*]}; do
+                if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then
+                    continue
+                fi
+                for threads in ${cpp_cpu_threads_list[*]}; do
+                    for batch_size in ${cpp_batch_size_list[*]}; do
+                        _save_log_path="${_log_path}/cpp_infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_batchsize_${batch_size}.log"
+                        set_infer_data=$(func_set_params "${cpp_image_dir_key}" "${_img_dir}")
+                        set_benchmark=$(func_set_params "${cpp_benchmark_key}" "${cpp_benchmark_value}")
+                        set_batchsize=$(func_set_params "${cpp_batch_size_key}" "${batch_size}")
+                        set_cpu_threads=$(func_set_params "${cpp_cpu_threads_key}" "${threads}")
+                        set_model_dir=$(func_set_params "${cpp_infer_model_key}" "${_model_dir}")
+                        command="${_script} ${cpp_use_gpu_key}=${use_gpu} ${cpp_use_mkldnn_key}=${use_mkldnn} ${set_cpu_threads} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} > ${_save_log_path} 2>&1 "
+                        eval $command
+                        last_status=${PIPESTATUS[0]}
+                        eval "cat ${_save_log_path}"
+                        status_check $last_status "${command}" "${status_log}"
+                    done
+                done
+            done
+        elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then
+            for use_trt in ${cpp_use_trt_list[*]}; do
+                for precision in ${cpp_precision_list[*]}; do
+                    if [[ ${_flag_quant} = "False" ]] && [[ ${precision} =~ "int8" ]]; then
+                        continue
+                    fi 
+                    if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then
+                        continue
+                    fi
+                    if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then
+                        continue
+                    fi
+                    for batch_size in ${cpp_batch_size_list[*]}; do
+                        _save_log_path="${_log_path}/cpp_infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log"
+                        set_infer_data=$(func_set_params "${cpp_image_dir_key}" "${_img_dir}")
+                        set_benchmark=$(func_set_params "${cpp_benchmark_key}" "${cpp_benchmark_value}")
+                        set_batchsize=$(func_set_params "${cpp_batch_size_key}" "${batch_size}")
+                        set_tensorrt=$(func_set_params "${cpp_use_trt_key}" "${use_trt}")
+                        set_precision=$(func_set_params "${cpp_precision_key}" "${precision}")
+                        set_model_dir=$(func_set_params "${cpp_infer_model_key}" "${_model_dir}")
+                        command="${_script} ${cpp_use_gpu_key}=${use_gpu} ${set_tensorrt} ${set_precision} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} > ${_save_log_path} 2>&1 "
+                        eval $command
+                        last_status=${PIPESTATUS[0]}
+                        eval "cat ${_save_log_path}"
+                        status_check $last_status "${command}" "${status_log}"
+                    done
+                done
+            done
+        else
+            echo "Does not support hardware other than CPU and GPU Currently!"
+        fi
+    done
+}
 if [ ${MODE} = "infer" ]; then
    GPUID=$3
    if [ ${#GPUID} -le 0 ];then
@@ -252,6 +344,25 @@ if [ ${MODE} = "infer" ]; then
        Count=$(($Count + 1))
    done
+elif [ ${MODE} = "cpp_infer" ]; then
+    GPUID=$3
+    if [ ${#GPUID} -le 0 ];then
+        env=" "
+    else
+        env="export CUDA_VISIBLE_DEVICES=${GPUID}"
+    fi
+    # set CUDA_VISIBLE_DEVICES
+    eval $env
+    export Count=0
+    IFS="|"
+    infer_quant_flag=(${cpp_infer_is_quant})
+    for infer_model in ${cpp_infer_model_dir_list[*]}; do
+        #run inference
+        is_quant=${infer_quant_flag[Count]}
+        func_cpp_inference "${inference_cmd}" "${infer_model}" "${LOG_PATH}" "${cpp_infer_img_dir}" ${is_quant}
+        Count=$(($Count + 1))
+    done
 else
    IFS="|"
    export Count=0

--- a/tools/infer/predict_det.py
+++ b/tools/infer/predict_det.py
@@ -109,6 +109,7 @@ class TextDetector(object):
        if args.benchmark:
            import auto_log
            pid = os.getpid()
+            gpu_id = utility.get_infer_gpuid()
            self.autolog = auto_log.AutoLogger(
                model_name="det",
                model_precision=args.precision,
@@ -118,7 +119,7 @@ class TextDetector(object):
                inference_config=self.config,
                pids=pid,
                process_name=None,
-                gpu_ids=0,
+                gpu_ids=gpu_id if args.use_gpu else None,
                time_keys=[
                    'preprocess_time', 'inference_time', 'postprocess_time'
                ],
@@ -216,7 +217,7 @@ class TextDetector(object):
            preds['f_score'] = outputs[1]
            preds['f_tco'] = outputs[2]
            preds['f_tvo'] = outputs[3]
-        elif self.det_algorithm in ['DB','PSE']:
+        elif self.det_algorithm in ['DB', 'PSE']:
            preds['maps'] = outputs[0]
        else:
            raise NotImplementedError
@@ -224,7 +225,9 @@ class TextDetector(object):
        #self.predictor.try_shrink_memory()
        post_result = self.postprocess_op(preds, shape_list)
        dt_boxes = post_result[0]['points']
-        if (self.det_algorithm == "SAST" and self.det_sast_polygon) or (self.det_algorithm == "PSE" and self.det_pse_box_type=='poly'):
+        if (self.det_algorithm == "SAST" and
+                self.det_sast_polygon) or (self.det_algorithm == "PSE" and
+                                           self.det_pse_box_type == 'poly'):
            dt_boxes = self.filter_tag_det_res_only_clip(dt_boxes, ori_im.shape)
        else:
            dt_boxes = self.filter_tag_det_res(dt_boxes, ori_im.shape)

--- a/tools/infer/predict_e2e.py
+++ b/tools/infer/predict_e2e.py
@@ -74,7 +74,7 @@ class TextE2E(object):
        self.preprocess_op = create_operators(pre_process_list)
        self.postprocess_op = build_post_process(postprocess_params)
-        self.predictor, self.input_tensor, self.output_tensors = utility.create_predictor(
+        self.predictor, self.input_tensor, self.output_tensors, _ = utility.create_predictor(
            args, 'e2e', logger)  # paddle.jit.load(args.det_model_dir)
        # self.predictor.eval()

--- a/tools/infer/predict_rec.py
+++ b/tools/infer/predict_rec.py
@@ -68,6 +68,7 @@ class TextRecognizer(object):
        if args.benchmark:
            import auto_log
            pid = os.getpid()
+            gpu_id = utility.get_infer_gpuid()
            self.autolog = auto_log.AutoLogger(
                model_name="rec",
                model_precision=args.precision,
@@ -77,7 +78,7 @@ class TextRecognizer(object):
                inference_config=self.config,
                pids=pid,
                process_name=None,
-                gpu_ids=0 if args.use_gpu else None,
+                gpu_ids=gpu_id if args.use_gpu else None,
                time_keys=[
                    'preprocess_time', 'inference_time', 'postprocess_time'
                ],
@@ -87,7 +88,7 @@ class TextRecognizer(object):
    def resize_norm_img(self, img, max_wh_ratio):
        imgC, imgH, imgW = self.rec_image_shape
        assert imgC == img.shape[2]
-        if self.character_type == "ch":
+        max_wh_ratio = max(max_wh_ratio, imgW / imgH)
        imgW = int((32 * max_wh_ratio))
        h, w = img.shape[:2]
        ratio = w / float(h)
@@ -277,7 +278,7 @@ def main(args):
    if args.warmup:
        img = np.random.uniform(0, 255, [32, 320, 3]).astype(np.uint8)
        for i in range(2):
-            res = text_recognizer([img])
+            res = text_recognizer([img] * int(args.rec_batch_num))
    for image_file in image_file_list:
        img, flag = check_and_read_gif(image_file)

--- a/tools/infer/utility.py
+++ b/tools/infer/utility.py
@@ -166,6 +166,11 @@ def create_predictor(args, mode, logger):
        precision = inference.PrecisionType.Float32
    if args.use_gpu:
+        gpu_id = get_infer_gpuid()
+        if gpu_id is None:
+            raise ValueError(
+                "Not found GPU in current device. Please check your device or set args.use_gpu as False"
+            )
        config.enable_use_gpu(args.gpu_mem, 0)
        if args.use_tensorrt:
            config.enable_tensorrt_engine(
@@ -287,6 +292,20 @@ def create_predictor(args, mode, logger):
    return predictor, input_tensor, output_tensors, config
+def get_infer_gpuid():
+    cmd = "nvidia-smi"
+    res = os.popen(cmd).readlines()
+    if len(res) == 0:
+        return None
+    cmd = "env | grep CUDA_VISIBLE_DEVICES"
+    env_cuda = os.popen(cmd).readlines()
+    if len(env_cuda) == 0:
+        return 0
+    else:
+        gpu_id = env_cuda[0].strip().split("=")[1]
+        return int(gpu_id[0])
 def draw_e2e_res(dt_boxes, strs, img_path):
    src_im = cv2.imread(img_path)
    for box, str in zip(dt_boxes, strs):

--- a/tools/program.py
+++ b/tools/program.py
@@ -186,6 +186,8 @@ def train(config,
    model.train()
    use_srn = config['Architecture']['algorithm'] == "SRN"
+    use_nrtr = config['Architecture']['algorithm'] == "NRTR"
    try:
        model_type = config['Architecture']['model_type']
    except:
@@ -213,7 +215,7 @@ def train(config,
            images = batch[0]
            if use_srn:
                model_average = True
-            if use_srn or model_type == 'table':
+            if use_srn or model_type == 'table' or use_nrtr:
                preds = model(images, data=batch[1:])
            else:
                preds = model(images)
@@ -398,7 +400,7 @@ def preprocess(is_train=False):
    alg = config['Architecture']['algorithm']
    assert alg in [
        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
-        'CLS', 'PGNet', 'Distillation', 'TableAttn', 'PSE'
+        'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn', 'PSE'
    ]
    device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'