Merge branch 'v0.4.0' of https://github.com/PaddlePaddle/Serving into v0.4.0

doc/DOCKER_IMAGES.md

@@ -25,7 +25,6 @@ You can get images in two ways:
 
 
 
-
 ## Image description
 
 Runtime images cannot be used for compilation.
@@ -40,3 +39,13 @@ Runtime images cannot be used for compilation.
 |              GPU (cuda10.0-cudnn7) development               | CentOS7 | latest-cuda10.0-cudnn7-devel | [Dockerfile.cuda10.0-cudnn7.devel](../tools/Dockerfile.cuda10.0-cudnn7.devel) |
 |     CPU development (Used to compile packages on Ubuntu)     | CentOS6 |            <None>            | [Dockerfile.centos6.devel](../tools/Dockerfile.centos6.devel) |
 | GPU (cuda9.0-cudnn7) development (Used to compile packages on Ubuntu) | CentOS6 |            <None>            | [Dockerfile.centos6.cuda9.0-cudnn7.devel](../tools/Dockerfile.centos6.cuda9.0-cudnn7.devel) |
+
+
+
+## Requirements for running CUDA containers
+
+Running a CUDA container requires a machine with at least one CUDA-capable GPU and a driver compatible with the CUDA toolkit version you are using. 
+
+The machine running the CUDA container **only requires the NVIDIA driver**, the CUDA toolkit doesn't have to be installed.
+
+For the relationship between CUDA toolkit version, Driver version and GPU architecture, please refer to [nvidia-docker wiki](https://github.com/NVIDIA/nvidia-docker/wiki/CUDA).

doc/DOCKER_IMAGES_CN.md

@@ -25,7 +25,6 @@
    
 
 
-
 ## 镜像说明
 
 运行时镜像不能用于开发编译。
@@ -40,3 +39,13 @@
 | GPU (cuda10.0-cudnn7) 开发镜像                     | CentOS7  | latest-cuda10.0-cudnn7-devel | [Dockerfile.cuda10.0-cudnn7.devel](../tools/Dockerfile.cuda10.0-cudnn7.devel) |
 | CPU 开发镜像 (用于编译 Ubuntu 包)                  | CentOS6  | <无>                         | [Dockerfile.centos6.devel](../tools/Dockerfile.centos6.devel) |
 | GPU (cuda9.0-cudnn7) 开发镜像 (用于编译 Ubuntu 包) | CentOS6  | <无>                         | [Dockerfile.centos6.cuda9.0-cudnn7.devel](../tools/Dockerfile.centos6.cuda9.0-cudnn7.devel) |
+
+
+
+## 运行CUDA容器的要求
+
+运行CUDA容器需要至少具有一个支持CUDA的GPU以及与您所使用的CUDA工具包版本兼容的驱动程序。
+
+运行CUDA容器的机器**只需要相应的NVIDIA驱动程序**，而CUDA工具包不是必要的。
+
+相关CUDA工具包版本、驱动版本和GPU架构的关系请参阅 [nvidia-docker wiki](https://github.com/NVIDIA/nvidia-docker/wiki/CUDA)。

doc/INFERENCE_TO_SERVING.md

@@ -24,13 +24,13 @@ inference_model_dir = "your_inference_model"
 serving_client_dir = "serving_client_dir"
 serving_server_dir = "serving_server_dir"
 feed_var_names, fetch_var_names = inference_model_to_serving(
-		inference_model_dir, serving_client_dir, serving_server_dir)
+		inference_model_dir, serving_server_dir, serving_client_dir)
 ```
 
 if your model file and params file are both standalone, please use the following api.
 
 ```
 feed_var_names, fetch_var_names = inference_model_to_serving(
-		inference_model_dir, serving_client_dir, serving_server_dir,
+		inference_model_dir, serving_server_dir, serving_client_dir,
 		model_filename="model", params_filename="params")
 ```

doc/INFERENCE_TO_SERVING_CN.md

@@ -23,11 +23,11 @@ inference_model_dir = "your_inference_model"
 serving_client_dir = "serving_client_dir"
 serving_server_dir = "serving_server_dir"
 feed_var_names, fetch_var_names = inference_model_to_serving(
-		inference_model_dir, serving_client_dir, serving_server_dir)
+		inference_model_dir, serving_server_dir, serving_client_dir)
 ```
 如果模型中有模型描述文件`model_filename` 和 模型参数文件`params_filename`，那么请用
 ```
 feed_var_names, fetch_var_names = inference_model_to_serving(
-		inference_model_dir, serving_client_dir, serving_server_dir,
+		inference_model_dir, serving_server_dir, serving_client_dir,
 		 model_filename="model", params_filename="params")
 ```

java/examples/pom.xml

@@ -75,7 +75,7 @@
         <dependency>
             <groupId>junit</groupId>
             <artifactId>junit</artifactId>
-            <version>4.11</version>
+            <version>4.13.1</version>
             <scope>test</scope>
         </dependency>
         <dependency>

python/examples/cascade_rcnn/test_client.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from paddle_serving_client import Client
+from paddle_serving_app.reader import *
+import numpy as np
+
+preprocess = Sequential([
+    File2Image(), BGR2RGB(), Div(255.0),
+    Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], False),
+    Resize(800, 1333), Transpose((2, 0, 1)), PadStride(32)
+])
+postprocess = RCNNPostprocess("label_list.txt", "output")
+client = Client()
+client.load_client_config("serving_client/serving_client_conf.prototxt")
+client.connect(['127.0.0.1:9292'])
+im = preprocess('000000570688.jpg')
+fetch_map = client.predict(
+    feed={
+        "image": im,
+        "im_info": np.array(list(im.shape[1:]) + [1.0]),
+        "im_shape": np.array(list(im.shape[1:]) + [1.0])
+    },
+    fetch=["multiclass_nms_0.tmp_0"],
+    batch=False)
+fetch_map["image"] = '000000570688.jpg'
+print(fetch_map)
+postprocess(fetch_map)
+print(fetch_map)

python/examples/faster_rcnn/benchmark.py

+# -*- coding: utf-8 -*-
+#
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# pylint: disable=doc-string-missing
+
+from __future__ import unicode_literals, absolute_import
+import os
+import sys
+import time
+import json
+import requests
+from paddle_serving_client import Client
+from paddle_serving_client.utils import MultiThreadRunner
+from paddle_serving_client.utils import benchmark_args, show_latency
+from paddle_serving_app.reader import ChineseBertReader
+
+from paddle_serving_app.reader import *
+import numpy as np
+
+args = benchmark_args()
+
+
+def single_func(idx, resource):
+    img = "./000000570688.jpg"
+    profile_flags = False
+    latency_flags = False
+    if os.getenv("FLAGS_profile_client"):
+        profile_flags = True
+    if os.getenv("FLAGS_serving_latency"):
+        latency_flags = True
+        latency_list = []
+
+    if args.request == "rpc":
+        preprocess = Sequential([
+            File2Image(), BGR2RGB(), Div(255.0),
+            Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], False),
+            Resize(640, 640), Transpose((2, 0, 1))
+        ])
+
+        postprocess = RCNNPostprocess("label_list.txt", "output")
+        client = Client()
+
+        client.load_client_config(args.model)
+        client.connect([resource["endpoint"][idx % len(resource["endpoint"])]])
+
+        start = time.time()
+        for i in range(turns):
+            if args.batch_size >= 1:
+                l_start = time.time()
+                feed_batch = []
+                b_start = time.time()
+                im = preprocess(img)
+                for bi in range(args.batch_size):
+                    print("1111batch")
+                    print(bi)
+                    feed_batch.append({
+                        "image": im,
+                        "im_info": np.array(list(im.shape[1:]) + [1.0]),
+                        "im_shape": np.array(list(im.shape[1:]) + [1.0])
+                    })
+            # im = preprocess(img)
+                b_end = time.time()
+
+                if profile_flags:
+                    sys.stderr.write(
+                        "PROFILE\tpid:{}\tbert_pre_0:{} bert_pre_1:{}\n".format(
+                            os.getpid(),
+                            int(round(b_start * 1000000)),
+                            int(round(b_end * 1000000))))
+                #result = client.predict(feed=feed_batch, fetch=fetch)
+                fetch_map = client.predict(
+                    feed=feed_batch, fetch=["multiclass_nms"])
+                fetch_map["image"] = img
+                postprocess(fetch_map)
+
+                l_end = time.time()
+                if latency_flags:
+                    latency_list.append(l_end * 1000 - l_start * 1000)
+            else:
+                print("unsupport batch size {}".format(args.batch_size))
+    else:
+        raise ValueError("not implemented {} request".format(args.request))
+    end = time.time()
+    if latency_flags:
+        return [[end - start], latency_list]
+    else:
+        return [[end - start]]
+
+
+if __name__ == '__main__':
+    multi_thread_runner = MultiThreadRunner()
+    endpoint_list = ["127.0.0.1:7777"]
+    turns = 10
+    start = time.time()
+    result = multi_thread_runner.run(
+        single_func, args.thread, {"endpoint": endpoint_list,
+                                   "turns": turns})
+    end = time.time()
+    total_cost = end - start
+
+    avg_cost = 0
+    for i in range(args.thread):
+        avg_cost += result[0][i]
+    avg_cost = avg_cost / args.thread
+
+    print("total cost: {}s".format(total_cost))
+    print("each thread cost: {}s. ".format(avg_cost))
+    print("qps: {}samples/s".format(args.batch_size * args.thread * turns /
+                                    total_cost))
+    if os.getenv("FLAGS_serving_latency"):
+        show_latency(result[1])

python/examples/faster_rcnn/benchmark.sh

+rm profile_log*
+export CUDA_VISIBLE_DEVICES=0
+export FLAGS_profile_server=1
+export FLAGS_profile_client=1
+export FLAGS_serving_latency=1
+
+gpu_id=0
+#save cpu and gpu utilization log
+if [ -d utilization ];then
+    rm -rf utilization
+else
+    mkdir utilization
+fi
+#start server
+$PYTHONROOT/bin/python3 -m paddle_serving_server_gpu.serve --model $1 --port 7777 --thread 4 --gpu_ids 0  --ir_optim >  elog  2>&1 &
+sleep 5
+
+#warm up
+$PYTHONROOT/bin/python3 benchmark.py --thread 4 --batch_size 1 --model $2/serving_client_conf.prototxt --request rpc > profile 2>&1
+echo -e "import psutil\ncpu_utilization=psutil.cpu_percent(1,False)\nprint('CPU_UTILIZATION:', cpu_utilization)\n" > cpu_utilization.py
+for thread_num in 1 4 8 16
+do
+for batch_size in 1
+do
+    job_bt=`date '+%Y%m%d%H%M%S'`
+    nvidia-smi --id=0 --query-compute-apps=used_memory --format=csv -lms 100 > gpu_use.log 2>&1 &
+    nvidia-smi --id=0 --query-gpu=utilization.gpu --format=csv -lms 100 > gpu_utilization.log 2>&1 &
+    gpu_memory_pid=$!
+    $PYTHONROOT/bin/python3 benchmark.py --thread $thread_num --batch_size $batch_size --model $2/serving_client_conf.prototxt --request rpc > profile 2>&1
+    kill ${gpu_memory_pid}
+    kill `ps -ef|grep used_memory|awk '{print $2}'`
+    echo "model_name:" $1
+    echo "thread_num:" $thread_num
+    echo "batch_size:" $batch_size
+    echo "=================Done===================="
+    echo "model_name:$1" >> profile_log_$1
+    echo "batch_size:$batch_size" >> profile_log_$1
+    $PYTHONROOT/bin/python3 cpu_utilization.py >> profile_log_$1
+    job_et=`date '+%Y%m%d%H%M%S'`
+    awk 'BEGIN {max = 0} {if(NR>1){if ($1 > max) max=$1}} END {print "MAX_GPU_MEMORY:", max}' gpu_use.log >> profile_log_$1
+    awk 'BEGIN {max = 0} {if(NR>1){if ($1 > max) max=$1}} END {print "GPU_UTILIZATION:", max}' gpu_utilization.log >> profile_log_$1
+    rm -rf gpu_use.log gpu_utilization.log
+    $PYTHONROOT/bin/python3 ../util/show_profile.py profile $thread_num >> profile_log_$1
+    tail -n 8 profile >> profile_log_$1
+    echo "" >> profile_log_$1
+done
+done
+
+#Divided log
+awk 'BEGIN{RS="\n\n"}{i++}{print > "bert_log_"i}' profile_log_$1
+mkdir bert_log && mv bert_log_* bert_log
+ps -ef|grep 'serving'|grep -v grep|cut -c 9-15 | xargs kill -9

python/examples/faster_rcnn/label_list.txt

+background
+person
+bicycle
+car
+motorcycle
+airplane
+bus
+train
+truck
+boat
+traffic light
+fire hydrant
+stop sign
+parking meter
+bench
+bird
+cat
+dog
+horse
+sheep
+cow
+elephant
+bear
+zebra
+giraffe
+backpack
+umbrella
+handbag
+tie
+suitcase
+frisbee
+skis
+snowboard
+sports ball
+kite
+baseball bat
+baseball glove
+skateboard
+surfboard
+tennis racket
+bottle
+wine glass
+cup
+fork
+knife
+spoon
+bowl
+banana
+apple
+sandwich
+orange
+broccoli
+carrot
+hot dog
+pizza
+donut
+cake
+chair
+couch
+potted plant
+bed
+dining table
+toilet
+tv
+laptop
+mouse
+remote
+keyboard
+cell phone
+microwave
+oven
+toaster
+sink
+refrigerator
+book
+clock
+vase
+scissors
+teddy bear
+hair drier
+toothbrush

python/examples/ocr/det_debugger_server.py

@@ -53,7 +53,9 @@ class OCRService(WebService):
         self.ori_h, self.ori_w, _ = im.shape
         det_img = self.det_preprocess(im)
         _, self.new_h, self.new_w = det_img.shape
-        return {"image": det_img[np.newaxis, :].copy()}, ["concat_1.tmp_0"]
+        return {
+            "image": det_img[np.newaxis, :].copy()
+        }, ["concat_1.tmp_0"], True
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         det_out = fetch_map["concat_1.tmp_0"]

python/examples/ocr/det_web_server.py

@@ -54,7 +54,7 @@ class OCRService(WebService):
         det_img = self.det_preprocess(im)
         _, self.new_h, self.new_w = det_img.shape
         print(det_img)
-        return {"image": det_img}, ["concat_1.tmp_0"]
+        return {"image": det_img}, ["concat_1.tmp_0"], False
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         det_out = fetch_map["concat_1.tmp_0"]

python/examples/ocr/ocr_debugger_server.py

@@ -42,10 +42,9 @@ class OCRService(WebService):
         self.det_client = LocalPredictor()
         if sys.argv[1] == 'gpu':
             self.det_client.load_model_config(
-                det_model_config, gpu=True, profile=False)
+                det_model_config, use_gpu=True, gpu_id=1)
         elif sys.argv[1] == 'cpu':
-            self.det_client.load_model_config(
-                det_model_config, gpu=False, profile=False)
+            self.det_client.load_model_config(det_model_config)
         self.ocr_reader = OCRReader()
 
     def preprocess(self, feed=[], fetch=[]):
@@ -58,7 +57,7 @@ class OCRService(WebService):
         det_img = det_img[np.newaxis, :]
         det_img = det_img.copy()
         det_out = self.det_client.predict(
-            feed={"image": det_img}, fetch=["concat_1.tmp_0"])
+            feed={"image": det_img}, fetch=["concat_1.tmp_0"], batch=True)
         filter_func = FilterBoxes(10, 10)
         post_func = DBPostProcess({
             "thresh": 0.3,
@@ -91,7 +90,7 @@ class OCRService(WebService):
             imgs[id] = norm_img
         feed = {"image": imgs.copy()}
         fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
-        return feed, fetch
+        return feed, fetch, True
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)
@@ -107,7 +106,8 @@ ocr_service.load_model_config("ocr_rec_model")
 ocr_service.prepare_server(workdir="workdir", port=9292)
 ocr_service.init_det_debugger(det_model_config="ocr_det_model")
 if sys.argv[1] == 'gpu':
-    ocr_service.run_debugger_service(gpu=True)
+    ocr_service.set_gpus("2")
+    ocr_service.run_debugger_service()
 elif sys.argv[1] == 'cpu':
     ocr_service.run_debugger_service()
 ocr_service.run_web_service()

python/examples/ocr/ocr_web_client.py

@@ -36,4 +36,5 @@ for img_file in os.listdir(test_img_dir):
     image = cv2_to_base64(image_data1)
     data = {"feed": [{"image": image}], "fetch": ["res"]}
     r = requests.post(url=url, headers=headers, data=json.dumps(data))
+    print(r)
     print(r.json())

python/examples/ocr/ocr_web_server.py

@@ -50,7 +50,7 @@ class OCRService(WebService):
         ori_h, ori_w, _ = im.shape
         det_img = self.det_preprocess(im)
         det_out = self.det_client.predict(
-            feed={"image": det_img}, fetch=["concat_1.tmp_0"])
+            feed={"image": det_img}, fetch=["concat_1.tmp_0"], batch=False)
         _, new_h, new_w = det_img.shape
         filter_func = FilterBoxes(10, 10)
         post_func = DBPostProcess({
@@ -77,10 +77,10 @@ class OCRService(WebService):
             max_wh_ratio = max(max_wh_ratio, wh_ratio)
         for img in img_list:
             norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
-            feed = {"image": norm_img}
-            feed_list.append(feed)
+            feed_list.append(norm_img[np.newaxis, :])
+        feed_batch = {"image": np.concatenate(feed_list, axis=0)}
         fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
-        return feed_list, fetch
+        return feed_batch, fetch, True
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)

python/examples/ocr/rec_debugger_server.py

@@ -52,7 +52,7 @@ class OCRService(WebService):
             imgs[i] = norm_img
         feed = {"image": imgs.copy()}
         fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
-        return feed, fetch
+        return feed, fetch, True
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)

python/examples/ocr/rec_web_server.py

@@ -51,10 +51,17 @@ class OCRService(WebService):
             max_wh_ratio = max(max_wh_ratio, wh_ratio)
         for img in img_list:
             norm_img = self.ocr_reader.resize_norm_img(img, max_wh_ratio)
-            feed = {"image": norm_img}
-            feed_list.append(feed)
+            #feed = {"image": norm_img}
+            feed_list.append(norm_img)
+        if len(feed_list) == 1:
+            feed_batch = {
+                "image": np.concatenate(
+                    feed_list, axis=0)[np.newaxis, :]
+            }
+        else:
+            feed_batch = {"image": np.concatenate(feed_list, axis=0)}
         fetch = ["ctc_greedy_decoder_0.tmp_0", "softmax_0.tmp_0"]
-        return feed_list, fetch
+        return feed_batch, fetch, True
 
     def postprocess(self, feed={}, fetch=[], fetch_map=None):
         rec_res = self.ocr_reader.postprocess(fetch_map, with_score=True)

python/paddle_serving_server/web_service.py

@@ -189,7 +189,7 @@ class WebService(object):
         from paddle_serving_app.local_predict import LocalPredictor
         self.client = LocalPredictor()
         self.client.load_model_config(
-            "{}".format(self.model_config), gpu=False, profile=False)
+            "{}".format(self.model_config), use_gpu=False)
 
     def run_web_service(self):
         print("This API will be deprecated later. Please do not use it")

python/paddle_serving_server_gpu/web_service.py

@@ -250,7 +250,7 @@ class WebService(object):
         from paddle_serving_app.local_predict import LocalPredictor
         self.client = LocalPredictor()
         self.client.load_model_config(
-            "{}".format(self.model_config), gpu=gpu, profile=False)
+            "{}".format(self.model_config), use_gpu=True, gpu_id=self.gpus[0])
 
     def run_web_service(self):
         print("This API will be deprecated later. Please do not use it")

python/setup.py.client.in

@@ -43,8 +43,8 @@ if '${PACK}' == 'ON':
     copy_lib()
 
 REQUIRED_PACKAGES = [
-    'six >= 1.10.0', 'protobuf >= 3.11.0', 'numpy >= 1.12', 'grpcio >= 1.28.1',
-    'grpcio-tools >= 1.28.1'
+    'six >= 1.10.0', 'protobuf >= 3.11.0', 'numpy >= 1.12', 'grpcio <= 1.33.2',
+    'grpcio-tools <= 1.33.2'
 ]
 
 

python/setup.py.server.in

@@ -28,7 +28,7 @@ max_version, mid_version, min_version = util.python_version()
 util.gen_pipeline_code("paddle_serving_server")
 
 REQUIRED_PACKAGES = [
-    'six >= 1.10.0', 'protobuf >= 3.11.0', 'grpcio >= 1.28.1', 'grpcio-tools >= 1.28.1',
+    'six >= 1.10.0', 'protobuf >= 3.11.0', 'grpcio <= 1.33.2', 'grpcio-tools <= 1.33.2',
     'paddle_serving_client', 'flask >= 1.1.1', 'paddle_serving_app', 'func_timeout', 'pyyaml'
 ]
 

python/setup.py.server_gpu.in

@@ -30,7 +30,7 @@ max_version, mid_version, min_version = util.python_version()
 util.gen_pipeline_code("paddle_serving_server_gpu")
 
 REQUIRED_PACKAGES = [
-    'six >= 1.10.0', 'protobuf >= 3.11.0', 'grpcio >= 1.28.1', 'grpcio-tools >= 1.28.1',
+    'six >= 1.10.0', 'protobuf >= 3.11.0', 'grpcio <= 1.33.2', 'grpcio-tools <= 1.33.2',
     'paddle_serving_client', 'flask >= 1.1.1', 'paddle_serving_app', 'func_timeout', 'pyyaml'
 ]
 

tools/Dockerfile.cuda10.1-cudnn7-trt6.devel

+FROM nvidia/cuda:10.1-cudnn7-devel-centos7
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" \
+    && yum -y install wget >/dev/null \
+    && yum -y install gcc gcc-c++ make glibc-static which  \
+    && yum -y install git openssl-devel curl-devel bzip2-devel python-devel \
+    && yum -y install libSM-1.2.2-2.el7.x86_64 --setopt=protected_multilib=false \
+    && yum -y install libXrender-0.9.10-1.el7.x86_64 --setopt=protected_multilib=false \
+    && yum -y install libXext-1.3.3-3.el7.x86_64 --setopt=protected_multilib=false 
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" && \
+    wget https://github.com/protocolbuffers/protobuf/releases/download/v3.11.2/protobuf-all-3.11.2.tar.gz && \
+    tar zxf protobuf-all-3.11.2.tar.gz && \
+    cd protobuf-3.11.2 && \
+    ./configure && make -j4 && make install && \
+    make clean && \
+    cd .. && rm -rf protobuf-*
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" && \
+    wget https://cmake.org/files/v3.2/cmake-3.2.0-Linux-x86_64.tar.gz >/dev/null \
+    && tar xzf cmake-3.2.0-Linux-x86_64.tar.gz \
+    && mv cmake-3.2.0-Linux-x86_64 /usr/local/cmake3.2.0 \
+    && echo 'export PATH=/usr/local/cmake3.2.0/bin:$PATH' >> /root/.bashrc \
+    && rm cmake-3.2.0-Linux-x86_64.tar.gz 
+
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" && \
+    wget https://dl.google.com/go/go1.14.linux-amd64.tar.gz >/dev/null \
+    && tar xzf go1.14.linux-amd64.tar.gz \
+    && mv go /usr/local/go \
+    && echo 'export GOROOT=/usr/local/go' >> /root/.bashrc \
+    && echo 'export PATH=/usr/local/go/bin:$PATH' >> /root/.bashrc \
+    && rm go1.14.linux-amd64.tar.gz 
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" && \
+    yum -y install python-devel sqlite-devel  \
+    && curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py >/dev/null \
+    && python get-pip.py >/dev/null \
+    && rm get-pip.py 
+
+RUN export http_proxy="http://172.19.56.199:3128" \
+    && export https_proxy="http://172.19.56.199:3128" && \
+    yum install -y python3 python3-devel \
+    && yum -y install epel-release && yum -y install patchelf libXext libSM libXrender\
+    && yum clean all 
+
+RUN localedef -c -i en_US -f UTF-8 en_US.UTF-8 \
+    && echo "export LANG=en_US.utf8" >> /root/.bashrc \
+    && echo "export LANGUAGE=en_US.utf8" >> /root/.bashrc
+
+RUN wget https://paddle-serving.bj.bcebos.com/tools/TensorRT-6.0.1.5.CentOS-7.6.x86_64-gnu.cuda-10.1.cudnn7.6.tar.gz \
+    && tar -xzf TensorRT-6.0.1.5.CentOS-7.6.x86_64-gnu.cuda-10.1.cudnn7.6.tar.gz \
+    && mv TensorRT-6.0.1.5 /usr/local/ \
+    && rm TensorRT-6.0.1.5.CentOS-7.6.x86_64-gnu.cuda-10.1.cudnn7.6.tar.gz \
+    && echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/TensorRT-6.0.1.5/lib/' >> /root/.bashrc