support k8s demo (Imagenet)

python/examples/imagenet/README.md

@@ -47,3 +47,13 @@ client send inference request
 python resnet50_rpc_client.py ResNet50_vd_client_config/serving_client_conf.prototxt
 ```
 *the port of server side in this example is 9696
+
+### Launch Paddle Serving on Kubernetes
+
+Paddle Serving support deployment on Kubernetes (K8S) clusters. From `imagenet_k8s_rpc.yaml` we define Serving as K8S Deployment and Service. User can deploy Serving on containers and expose internal or external service.
+
+We strongly recommend [Baidu Cloud CCE Cluster](https://cloud.baidu.com/search.html?q=CCE)
+
+```
+kubectl apply -f imagenet_k8s_rpc.yaml
+```

python/examples/imagenet/README_CN.md

@@ -47,3 +47,13 @@ client端进行预测
 python resnet50_rpc_client.py ResNet50_vd_client_config/serving_client_conf.prototxt
 ```
 *server端示例中服务端口为9696端口
+
+### K8S启动
+
+还可以运用K8S启动，在`imagenet_k8s_rpc.yaml`中定义了Serving的Deployment和Service，用户可以在K8S集群上启动Deployment并且通过Service对外暴露服务，用户可以在此基础上进行二次开发。
+
+推荐百度云的[CCE（K8S）集群](https://cloud.baidu.com/search.html?q=CCE)
+
+```
+kubectl apply -f imagenet_k8s_rpc.yaml
+```

python/examples/imagenet/imagenet_k8s_rpc.yaml

@@ -0,0 +1,55 @@
+# 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.
+
+apiVersion: apps/v1beta1
+kind: Deployment
+metadata:
+  name: paddleserving
+  labels:
+    app: paddleserving
+spec:
+  replicas: 1
+  template:
+    metadata:
+      name: paddleserving
+      labels:
+        app: paddleserving
+    spec:
+      containers:
+      - name: paddleserving
+        image: hub.baidubce.com/paddlepaddle/serving:latest
+        imagePullPolicy: Always
+        workingDir: /
+        command: ['/bin/bash', '-c']
+        args: ['pip install -U paddle-serving-server paddle-serving-client paddle-serving-app && \
+                python -m paddle_serving_app.package --get_model resnet_v2_50_imagenet && \
+                tar xf resnet_v2_50_imagenet.tar.gz && \
+                python -m paddle_serving_server.serve --model resnet_v2_50_imagenet_model/ --port 9696']
+        ports: 
+        - containerPort: 9696
+          name: serving
+
+---
+
+apiVersion: v1
+kind: Service
+metadata:
+  name: paddleserving
+spec:
+  ports:
+    - name: paddleserving
+      port: 9696
+      targetPort: 9696
+  selector:
+    app: paddleserving

python/examples/pipeline/faster_rcnn/benchmark.py

@@ -0,0 +1,107 @@
+import sys
+import os
+import yaml
+import requests
+import time
+import json
+import cv2
+import base64
+try:
+    from paddle_serving_server_gpu.pipeline import PipelineClient
+except ImportError:
+    from paddle_serving_server.pipeline import PipelineClient
+import numpy as np
+from paddle_serving_client.utils import MultiThreadRunner
+from paddle_serving_client.utils import benchmark_args, show_latency
+
+def cv2_to_base64(image):
+    return base64.b64encode(image).decode('utf8')
+
+def parse_benchmark(filein, fileout):
+    with open(filein, "r") as fin:
+        res = yaml.load(fin)
+        del_list = []
+        for key in res["DAG"].keys():
+            if "call" in key:
+                del_list.append(key)
+        for key in del_list:
+            del res["DAG"][key]
+    with open(fileout, "w") as fout:
+        yaml.dump(res, fout, default_flow_style=False)
+
+def gen_yml(device):
+    fin = open("config.yml", "r")
+    config = yaml.load(fin)
+    fin.close()
+    config["dag"]["tracer"] = {"interval_s": 10}
+    if device == "gpu":
+        config["op"]["bert"]["local_service_conf"]["device_type"] = 1
+        config["op"]["bert"]["local_service_conf"]["devices"] = "2"        
+    with open("config2.yml", "w") as fout: 
+        yaml.dump(config, fout, default_flow_style=False)
+
+def run_http(idx, batch_size):
+    print("start thread ({})".format(idx))
+    url = "http://127.0.0.1:18082/faster_rcnn/prediction"
+    with open(os.path.join(".", "000000570688.jpg"), 'rb') as file:
+        image_data1 = file.read()
+    image = cv2_to_base64(image_data1)
+
+    start = time.time()
+    for i in range(10):
+        data = {"key": [], "value": []}
+        for j in range(batch_size):
+            data["key"].append("image_" + str(j))
+            data["value"].append(image)
+        r = requests.post(url=url, data=json.dumps(data))
+        print("done")
+    end = time.time()
+    return [[end - start]]
+
+def multithread_http(thread, batch_size):
+    multi_thread_runner = MultiThreadRunner()
+    result = multi_thread_runner.run(run_http , thread, batch_size)
+
+def run_rpc(thread, batch_size):
+    client = PipelineClient()
+    client.connect(['127.0.0.1:9998'])
+    with open("data-c.txt", 'r') as fin:
+        start = time.time()
+        lines = fin.readlines()
+        start_idx = 0
+        while start_idx < len(lines):
+            end_idx = min(len(lines), start_idx + batch_size)
+            feed = {}
+            for i in range(start_idx, end_idx):
+                feed[str(i - start_idx)] = lines[i]
+            ret = client.predict(feed_dict=feed, fetch=["res"])
+            start_idx += batch_size
+            if start_idx > 1000:
+                break
+        end = time.time()
+    return [[end - start]]
+
+
+def multithread_rpc(thraed, batch_size):
+    multi_thread_runner = MultiThreadRunner()
+    result = multi_thread_runner.run(run_rpc , thread, batch_size)
+
+if __name__ == "__main__":
+    if sys.argv[1] == "yaml":
+        mode = sys.argv[2] # brpc/  local predictor
+        thread = int(sys.argv[3])
+        device = sys.argv[4]
+        gen_yml(device)
+    elif sys.argv[1] == "run":
+        mode = sys.argv[2] # http/ rpc
+        thread = int(sys.argv[3])
+        batch_size = int(sys.argv[4])
+        if mode == "http":
+            multithread_http(thread, batch_size)
+        elif mode == "rpc":
+            multithread_rpc(thread, batch_size)
+    elif sys.argv[1] == "dump":
+        filein = sys.argv[2]
+        fileout = sys.argv[3]
+        parse_benchmark(filein, fileout)
+

python/examples/pipeline/faster_rcnn/benchmark.sh

@@ -0,0 +1,60 @@
+export FLAGS_profile_pipeline=1
+alias python3="python3.6"
+modelname="bert"
+# HTTP
+ps -ef | grep web_service | awk '{print $2}' | xargs kill -9 
+sleep 3
+python3 benchmark.py yaml local_predictor 1 cpu 
+rm -rf profile_log_$modelname
+for thread_num in 1  
+do
+  for batch_size in 1 2
+  do
+    echo "----FasterRCNN thread num: $thread_num batch size: $batch_size mode:http ----" >>profile_log_$modelname
+    rm -rf PipelineServingLogs
+    rm -rf cpu_utilization.py
+    python3 web_service.py >web.log 2>&1 &
+    sleep 3
+    nvidia-smi --id=2 --query-compute-apps=used_memory --format=csv -lms 100 > gpu_use.log 2>&1 &
+    nvidia-smi --id=2 --query-gpu=utilization.gpu --format=csv -lms 100 > gpu_utilization.log 2>&1 &
+    echo "import psutil\ncpu_utilization=psutil.cpu_percent(1,False)\nprint('CPU_UTILIZATION:', cpu_utilization)\n" > cpu_utilization.py
+    python3 benchmark.py run http $thread_num $batch_size
+    python3 cpu_utilization.py >>profile_log_$modelname
+    ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+    python3 benchmark.py dump benchmark.log benchmark.tmp
+    mv benchmark.tmp benchmark.log
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "MAX_GPU_MEMORY:", max}' gpu_use.log >> profile_log_$modelname
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "GPU_UTILIZATION:", max}' gpu_utilization.log >> profile_log_$modelname
+    cat benchmark.log >> profile_log_$modelname
+    #rm -rf gpu_use.log gpu_utilization.log
+  done
+done
+# RPC
+exit
+ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+sleep 3
+python3 benchmark.py yaml local_predictor 1 gpu
+
+for thread_num in 1 8 16
+do
+  for batch_size in 1 10 100
+  do
+    echo "----Bert thread num: $thread_num batch size: $batch_size mode:rpc ----" >>profile_log_$modelname
+    rm -rf PipelineServingLogs
+    rm -rf cpu_utilization.py
+    python3 web_service.py >web.log 2>&1 &
+    sleep 3
+    nvidia-smi --id=2 --query-compute-apps=used_memory --format=csv -lms 100 > gpu_use.log 2>&1 &
+    nvidia-smi --id=2 --query-gpu=utilization.gpu --format=csv -lms 100 > gpu_utilization.log 2>&1 &
+    echo "import psutil\ncpu_utilization=psutil.cpu_percent(1,False)\nprint('CPU_UTILIZATION:', cpu_utilization)\n" > cpu_utilization.py
+    python3 benchmark.py run rpc $thread_num $batch_size
+    python3 cpu_utilization.py >>profile_log_$modelname
+    ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+    python3 benchmark.py dump benchmark.log benchmark.tmp
+    mv benchmark.tmp benchmark.log
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "MAX_GPU_MEMORY:", max}' gpu_use.log >> profile_log_$modelname
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "GPU_UTILIZATION:", max}' gpu_utilization.log >> profile_log_$modelname
+    #rm -rf gpu_use.log gpu_utilization.log
+    cat benchmark.log >> profile_log_$modelname
+  done
+done

python/examples/pipeline/faster_rcnn/config.yml

@@ -0,0 +1,17 @@
+dag:
+  is_thread_op: false
+  tracer:
+    interval_s: 10
+http_port: 18082
+op:
+  faster_rcnn:
+    local_service_conf:
+      client_type: local_predictor
+      concurrency: 2
+      device_type: 1
+      devices: '2'
+      fetch_list:
+      - save_infer_model/scale_0.tmp_1
+      model_config: serving_server/
+rpc_port: 9998
+worker_num: 20

python/examples/pipeline/faster_rcnn/label_list.txt

@@ -0,0 +1,80 @@
+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/pipeline/faster_rcnn/pipeline_http_client.py

@@ -0,0 +1,35 @@
+# 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_server.pipeline import PipelineClient
+import numpy as np
+import requests
+import json
+import cv2
+import base64
+import os
+
+
+def cv2_to_base64(image):
+    return base64.b64encode(image).decode('utf8')
+
+
+url = "http://127.0.0.1:18082/faster_rcnn/prediction"
+with open(os.path.join(".", "000000570688.jpg"), 'rb') as file:
+    image_data1 = file.read()
+image = cv2_to_base64(image_data1)
+
+for i in range(1):
+    data = {"key": ["image"], "value": [image]}
+    r = requests.post(url=url, data=json.dumps(data))
+    print(r.json())