Merge branch 'develop' into encryption

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/bert/bert_client.py

@@ -33,5 +33,5 @@ for line in sys.stdin:
     for key in feed_dict.keys():
         feed_dict[key] = np.array(feed_dict[key]).reshape((128, 1))
     #print(feed_dict)
-    result = client.predict(feed=feed_dict, fetch=fetch, batch=True)
+    result = client.predict(feed=feed_dict, fetch=fetch, batch=False)
 print(result)

python/examples/bert/bert_web_service.py

@@ -29,7 +29,7 @@ class BertService(WebService):
 
     def preprocess(self, feed=[], fetch=[]):
         feed_res = []
-        is_batch = True
+        is_batch = False
         for ins in feed:
             feed_dict = self.reader.process(ins["words"].encode("utf-8"))
             for key in feed_dict.keys():

python/examples/faster_rcnn_model/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_model/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/imdb/text_classify_service.py

@@ -29,13 +29,14 @@ class IMDBService(WebService):
     def preprocess(self, feed={}, fetch=[]):
         feed_batch = []
         words_lod = [0]
+        is_batch = True
         for ins in feed:
             words = self.dataset.get_words_only(ins["words"])
             words = np.array(words).reshape(len(words), 1)
             words_lod.append(words_lod[-1] + len(words))
             feed_batch.append(words)
         feed = {"words": np.concatenate(feed_batch), "words.lod": words_lod}
-        return feed, fetch
+        return feed, fetch, is_batch
 
 
 imdb_service = IMDBService(name="imdb")

python/paddle_serving_server/__init__.py

@@ -23,13 +23,13 @@ import paddle_serving_server as paddle_serving_server
 from .version import serving_server_version
 from contextlib import closing
 import collections
-import fcntl
-
 import shutil
 import numpy as np
 import grpc
 from .proto import multi_lang_general_model_service_pb2
 import sys
+if sys.platform.startswith('win') is False:
+    import fcntl
 sys.path.append(
     os.path.join(os.path.abspath(os.path.dirname(__file__)), 'proto'))
 from .proto import multi_lang_general_model_service_pb2_grpc

python/setup.py.app.in

@@ -32,8 +32,8 @@ if '${PACK}' == 'ON':
 
 
 REQUIRED_PACKAGES = [
-    'six >= 1.10.0', 'sentencepiece', 'opencv-python<=4.2.0.32', 'pillow',
-    'shapely<=1.6.1', 'pyclipper'
+    'six >= 1.10.0', 'sentencepiece<=0.1.92', 'opencv-python<=4.2.0.32', 'pillow',
+    'pyclipper'
 ]
 
 packages=['paddle_serving_app',

requirements.txt

+sphinx==2.1.0
+mistune
+sphinx_rtd_theme
+paddlepaddle>=1.8.4
+shapely<=1.6.1

doc/requirements.txt → requirements_win.txt

 sphinx==2.1.0
 mistune
 sphinx_rtd_theme
-paddlepaddle>=1.6
+paddlepaddle>=1.8.4
+shapely

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