Merge pull request #408 from MRXLT/0.2.0-cherry-v2

[cherry pick to 0.2.0 ] PR 406 407

go/serving_client/serving_client_api.go

@@ -27,7 +27,8 @@ import (
 type Tensor struct {
      Data   []byte `json:"data"`
      FloatData	   []float32 `json:"float_data"`
-     IntData	   []int64 `json:"int_data"`
+     IntData	   []int `json:"int_data"`
+     Int64Data	   []int64 `json:"int64_data"`
      ElemType	int `json:"elem_type"`
      Shape	[]int `json:"shape"`
 }
@@ -116,9 +117,9 @@ func Predict(handle Handle, int_feed_map map[string][]int64, fetch []string) map
      for i := 0; i < len(handle.FeedAliasNames); i++ {
      	 key_i := handle.FeedAliasNames[i]
 	 var tmp Tensor
-	 tmp.IntData = []int64{}
+	 tmp.IntData = []int{}
 	 tmp.Shape = []int{}
-	 tmp.IntData = int_feed_map[key_i]
+	 tmp.Int64Data = int_feed_map[key_i]
 	 tmp.ElemType = 0
 	 tmp.Shape = handle.FeedShapeMap[key_i]
 	 tensor_array = append(tensor_array, tmp)

python/examples/imdb/benchmark_batch.py

@@ -64,4 +64,4 @@ result = multi_thread_runner.run(single_func, args.thread, {})
 avg_cost = 0
 for cost in result[0]:
     avg_cost += cost
-print("total cost of each thread".format(avg_cost / args.thread))
+print("total cost {} s of each thread".format(avg_cost / args.thread))

python/paddle_serving_server_gpu/__init__.py

@@ -116,7 +116,7 @@ class Server(object):
         self.reload_interval_s = 10
         self.module_path = os.path.dirname(paddle_serving_server.__file__)
         self.cur_path = os.getcwd()
-        self.check_cuda
+        self.check_cuda()
         self.use_local_bin = False
         self.gpuid = 0
 
@@ -144,7 +144,7 @@ class Server(object):
             self.bin_path = os.environ["SERVING_BIN"]
 
     def check_cuda(self):
-        r = os.system("whereis cuda")
+        r = os.system("nvcc --version > /dev/null")
         if r != 0:
             raise SystemExit(
                 "CUDA not found, please check your environment or use cpu version by \"pip install paddle_serving_server\""