temp commit

python/examples/imagenet/benchmark.py

+# -*- coding: utf-8 -*-
+#
 # Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -13,13 +15,16 @@
 # limitations under the License.
 # pylint: disable=doc-string-missing
 
+from __future__ import unicode_literals, absolute_import
+import os
 import sys
-from image_reader import ImageReader
+import time
 from paddle_serving_client import Client
 from paddle_serving_client.utils import MultiThreadRunner
 from paddle_serving_client.utils import benchmark_args
-import time
-import os
+import requests
+import json
+from image_reader import ImageReader
 
 args = benchmark_args()
 
@@ -37,24 +42,31 @@ def single_func(idx, resource):
         client = Client()
         client.load_client_config(args.model)
         client.connect([resource["endpoint"][idx % len(resource["endpoint"])]])
-
         start = time.time()
         for i in range(1000):
-            img = reader.process_image(img_list[i]).reshape(-1)
-            fetch_map = client.predict(feed={"image": img}, fetch=["score"])
-        end = time.time()
-        return [[end - start]]
+            if args.batch_size >= 1:
+                feed_batch = []
+                for bi in range(args.batch_size):
+                    img = reader.process_image(img_list[i])
+                    img = img.reshape(-1)
+                    feed_batch.append({"image": img})
+                result = client.predict(feed=feed_batch, fetch=fetch)
+            else:
+                print("unsupport batch size {}".format(args.batch_size))
+
+    elif args.request == "http":
+        raise ("no batch predict for http")
+    end = time.time()
     return [[end - start]]
 
 
-if __name__ == "__main__":
+if __name__ == '__main__':
     multi_thread_runner = MultiThreadRunner()
     endpoint_list = ["127.0.0.1:9393"]
-    #card_num = 4
-    #for i in range(args.thread):
-    #    endpoint_list.append("127.0.0.1:{}".format(9295 + i % card_num))
+    #endpoint_list = endpoint_list + endpoint_list + endpoint_list
     result = multi_thread_runner.run(single_func, args.thread,
                                      {"endpoint": endpoint_list})
+    #result = single_func(0, {"endpoint": endpoint_list})
     avg_cost = 0
     for i in range(args.thread):
         avg_cost += result[0][i]

python/examples/imagenet/benchmark.sh

 rm profile_log
-for thread_num in 1 2 4 8 16
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+export FLAGS_profile_server=1
+export FLAGS_profile_client=1
+python -m paddle_serving_server_gpu.serve --model $1 --port 9292 --thread 4 --gpu_ids 0,1,2,3 2> elog > stdlog &
+
+sleep 5
+
+#warm up
+$PYTHONROOT/bin/python benchmark.py --thread 8 --batch_size 1 --model $2/serving_client_conf.prototxt --request rpc > profile 2>&1
+
+for thread_num in 4 8 16
 do
-    $PYTHONROOT/bin/python benchmark.py --thread $thread_num --model ResNet101_vd_client_config/serving_client_conf.prototxt --request rpc > profile 2>&1
-    echo "========================================"
-    echo "batch size : $batch_size" >> profile_log
+for batch_size in 1 4 16 64 256
+do
+    $PYTHONROOT/bin/python benchmark.py --thread $thread_num --batch_size $batch_size --model $2/serving_client_conf.prototxt --request rpc > profile 2>&1
+    echo "model name :" $1
+    echo "thread num :" $thread_num
+    echo "batch size :" $batch_size
+    echo "=================Done===================="
+    echo "model name :$1" >> profile_log
+    echo "batch size :$batch_size" >> profile_log
     $PYTHONROOT/bin/python ../util/show_profile.py profile $thread_num >> profile_log
-    tail -n 1 profile >> profile_log
+    tail -n 8 profile >> profile_log
+done
 done
+
+ps -ef|grep 'serving'|grep -v grep|cut -c 9-15 | xargs kill -9

python/examples/imagenet/benchmark_batch.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
-from paddle_serving_client import Client
-from paddle_serving_client.utils import MultiThreadRunner
-from paddle_serving_client.utils import benchmark_args
-import requests
-import json
-from image_reader import ImageReader
-
-args = benchmark_args()
-
-
-def single_func(idx, resource):
-    file_list = []
-    for file_name in os.listdir("./image_data/n01440764"):
-        file_list.append(file_name)
-    img_list = []
-    for i in range(1000):
-        img_list.append(open("./image_data/n01440764/" + file_list[i]).read())
-    if args.request == "rpc":
-        reader = ImageReader()
-        fetch = ["score"]
-        client = Client()
-        client.load_client_config(args.model)
-        client.connect([resource["endpoint"][idx % len(resource["endpoint"])]])
-        start = time.time()
-        for i in range(1000):
-            if args.batch_size >= 1:
-                feed_batch = []
-                for bi in range(args.batch_size):
-                    img = reader.process_image(img_list[i])
-                    img = img.reshape(-1)
-                    feed_batch.append({"image": img})
-                result = client.predict(feed=feed_batch, fetch=fetch)
-            else:
-                print("unsupport batch size {}".format(args.batch_size))
-
-    elif args.request == "http":
-        raise ("no batch predict for http")
-    end = time.time()
-    return [[end - start]]
-
-
-if __name__ == '__main__':
-    multi_thread_runner = MultiThreadRunner()
-    endpoint_list = ["127.0.0.1:9393"]
-    #endpoint_list = endpoint_list + endpoint_list + endpoint_list
-    result = multi_thread_runner.run(single_func, args.thread,
-                                     {"endpoint": endpoint_list})
-    #result = single_func(0, {"endpoint": endpoint_list})
-    avg_cost = 0
-    for i in range(args.thread):
-        avg_cost += result[0][i]
-    avg_cost = avg_cost / args.thread
-    print("average total cost {} s.".format(avg_cost))

python/examples/imagenet/benchmark_batch.sh

-rm profile_log
-for thread_num in 1 2 4 8 16
-do
-for batch_size in 1 2 4 8 16 32 64 128 256 512
-do
-    $PYTHONROOT/bin/python benchmark_batch.py --thread $thread_num --batch_size $batch_size --model ResNet101_vd_client_config/serving_client_conf.prototxt --request rpc > profile 2>&1
-    echo "========================================"
-    echo "batch size : $batch_size" >> profile_log
-    $PYTHONROOT/bin/python ../util/show_profile.py profile $thread_num >> profile_log
-    tail -n 1 profile >> profile_log
-done
-done

python/paddle_serving_app/models/model_list.py

@@ -84,6 +84,22 @@ class ServingModels(object):
             self.model_dict[
                 key] = "https://paddle-serving.bj.bcebos.com/paddle_hub_models/image/ImageClassification/" + key + ".tar.gz"
 
+        #SemanticModel
+        for key in [
+                "bert_cased_L-12_H-768_A-12", "bert_cased_L-24_H-1024_A-12",
+                "bert_chinese_L-12_H-768_A-12",
+                "bert_multi_cased_L-12_H-768_A-12",
+                "bert_multi_uncased_L-12_H-768_A-12",
+                "bert_uncased_L-12_H-768_A-12", "bert_uncased_L-24_H-1024_A-16",
+                "chinese-bert-wwm-ext", "chinese-bert-wwm",
+                "chinese-electra-base", "chinese-electra-small",
+                "chinese-electra-small", "chinese-roberta-wwm-ext", "ernie",
+                "ernie_tiny", "ernie_v2_eng_base", "ernie_v2_eng_large", "rbt3",
+                "rbtl3", "simnet_bow", "word2vec_skipgram"
+        ]:
+            self.model_dict[
+                key] = "https://paddle-serving.bj.bcebos.com/paddle_hub_models/text/SemanticModel/" + key + ".tar.gz"
+
     def get_model_list(self):
         return (self.model_dict.keys())