Merge pull request #974 from wangjiawei04/v0.4.1

cherry-pick 971 970 967

core/pdcodegen/CMakeLists.txt

@@ -7,6 +7,7 @@ PROTOBUF_GENERATE_CPP(pdcodegen_proto_srcs pdcodegen_proto_hdrs
 LIST(APPEND pdcodegen_srcs ${pdcodegen_proto_srcs})
 
 add_executable(pdcodegen ${pdcodegen_srcs})
+add_dependencies(pdcodegen boost)
 target_link_libraries(pdcodegen protobuf ${PROTOBUF_PROTOC_LIBRARY})
 
 # install

doc/LATEST_PACKAGES.md

@@ -19,7 +19,9 @@ https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post9-p
 #cuda 10.0
 https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post10-py3-none-any.whl
 #cuda10.1 with TensorRT 6
-https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.trt-py3-none-any.whl
+https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post101-py3-none-any.whl
+#cuda10.2 with TensorRT 7
+https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post102-py3-none-any.whl
 ```
 ### Python 2
 ```
@@ -28,7 +30,9 @@ https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post9-p
 #cuda 10.0
 https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post10-py2-none-any.whl
 ##cuda10.1 with TensorRT 6
-https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.trt-py2-none-any.whl
+https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post101-py2-none-any.whl
+#cuda10.2 with TensorRT 7
+https://paddle-serving.bj.bcebos.com/whl/paddle_serving_server_gpu-0.0.0.post102-py2-none-any.whl
 ```
 
 ## Client

java/examples/src/main/java/PipelineClientExample.java

@@ -62,7 +62,7 @@ public class PipelineClientExample {
                 return false;
             }
         }
-        PipelineFuture future = StaticPipelineClient.client.asyn_pr::qedict(feed_data, fetch,false,0);
+        PipelineFuture future = StaticPipelineClient.client.asyn_predict(feed_data, fetch,false,0);
         HashMap<String,String> result = future.get();
         if (result == null) {
             return false;

java/examples/src/main/java/StaticPipelineClient.java

@@ -37,7 +37,7 @@ public class StaticPipelineClient {
             System.out.println("already connect.");
             return true;
         }
-        succ = clieint.connect(target);
+        succ = client.connect(target);
         if (succ != true) {
             System.out.println("connect failed.");
             return false;

python/CMakeLists.txt

@@ -81,12 +81,20 @@ if (SERVER)
             DEPENDS ${SERVING_SERVER_CORE} server_config_py_proto ${PY_FILES})
         add_custom_target(paddle_python ALL DEPENDS ${PADDLE_SERVING_BINARY_DIR}/.timestamp)
     elseif(WITH_TRT)
+        if(CUDA_VERSION EQUAL 10.1)
+            set(SUFFIX 101)
+        elseif(CUDA_VERSION EQUAL 10.2)
+            set(SUFFIX 102)
+        elseif(CUDA_VERSION EQUAL 11.0)
+            set(SUFFIX 110)
+
+        endif()
         add_custom_command(
             OUTPUT ${PADDLE_SERVING_BINARY_DIR}/.timestamp
             COMMAND cp -r
             ${CMAKE_CURRENT_SOURCE_DIR}/paddle_serving_server_gpu/ ${PADDLE_SERVING_BINARY_DIR}/python/
             COMMAND env ${py_env} ${PYTHON_EXECUTABLE} gen_version.py
-            "server_gpu" trt
+            "server_gpu"  ${SUFFIX}
             COMMAND env ${py_env} ${PYTHON_EXECUTABLE} setup.py bdist_wheel
             DEPENDS ${SERVING_SERVER_CORE} server_config_py_proto ${PY_FILES})
         add_custom_target(paddle_python ALL DEPENDS ${PADDLE_SERVING_BINARY_DIR}/.timestamp)

python/paddle_serving_server/web_service.py

@@ -20,7 +20,7 @@ from paddle_serving_server import OpMaker, OpSeqMaker, Server
 from paddle_serving_client import Client
 from contextlib import closing
 import socket
-
+import numpy as np
 from paddle_serving_server import pipeline
 from paddle_serving_server.pipeline import Op
 
@@ -64,8 +64,8 @@ class WebService(object):
         f = open(client_config, 'r')
         model_conf = google.protobuf.text_format.Merge(
             str(f.read()), model_conf)
-        self.feed_names = [var.alias_name for var in model_conf.feed_var]
-        self.fetch_names = [var.alias_name for var in model_conf.fetch_var]
+        self.feed_vars = {var.name: var for var in model_conf.feed_var}
+        self.fetch_vars = {var.name: var for var in model_conf.fetch_var}
 
     def _launch_rpc_service(self):
         op_maker = OpMaker()
@@ -201,6 +201,15 @@ class WebService(object):
     def preprocess(self, feed=[], fetch=[]):
         print("This API will be deprecated later. Please do not use it")
         is_batch = True
+        feed_dict = {}
+        for var_name in self.feed_vars.keys():
+            feed_dict[var_name] = []
+        for feed_ins in feed:
+            for key in feed_ins:
+                feed_dict[key].append(np.array(feed_ins[key]).reshape(list(self.feed_vars[key].shape))[np.newaxis,:])
+        feed = {}
+        for key in feed_dict:
+            feed[key] = np.concatenate(feed_dict[key], axis=0) 
         return feed, fetch, is_batch
 
     def postprocess(self, feed=[], fetch=[], fetch_map=None):

python/paddle_serving_server_gpu/__init__.py

@@ -423,7 +423,7 @@ class Server(object):
         for line in version_file.readlines():
             if re.match("cuda_version", line):
                 cuda_version = line.split("\"")[1]
-                if cuda_version == "trt":
+                if cuda_version == "101" or cuda_version == "102" or cuda_version == "110":
                     device_version = "serving-gpu-" + cuda_version + "-"
                 elif cuda_version == "arm":
                     device_version = "serving-" + cuda_version + "-"

python/paddle_serving_server_gpu/web_service.py

@@ -70,8 +70,8 @@ class WebService(object):
         f = open(client_config, 'r')
         model_conf = google.protobuf.text_format.Merge(
             str(f.read()), model_conf)
-        self.feed_names = [var.alias_name for var in model_conf.feed_var]
-        self.fetch_names = [var.alias_name for var in model_conf.fetch_var]
+        self.feed_vars = {var.name: var for var in model_conf.feed_var}
+        self.fetch_vars = {var.name: var for var in model_conf.fetch_var}
 
     def set_gpus(self, gpus):
         print("This API will be deprecated later. Please do not use it")
@@ -278,6 +278,15 @@ class WebService(object):
     def preprocess(self, feed=[], fetch=[]):
         print("This API will be deprecated later. Please do not use it")
         is_batch = True
+        feed_dict = {}
+        for var_name in self.feed_vars.keys():
+            feed_dict[var_name] = []
+        for feed_ins in feed:
+            for key in feed_ins:
+                feed_dict[key].append(np.array(feed_ins[key]).reshape(list(self.feed_vars[key].shape))[np.newaxis,:])
+        feed = {}
+        for key in feed_dict:
+            feed[key] = np.concatenate(feed_dict[key], axis=0)
         return feed, fetch, is_batch
 
     def postprocess(self, feed=[], fetch=[], fetch_map=None):