add async

+import os
 import time
+import random
 from locust import HttpUser, task, between, constant, tag
 
 
+base_dir = '/home/lk/MyProject/BMW_F3OCR/数据集/文件分类/营业执照'
+file_path_list = [os.path.join(base_dir, file_name) for file_name in os.listdir(base_dir)]
+
+
 class QuickstartUser(HttpUser):
+    
     # wait_time = between(1, 5)
 
     @tag('sync')
@@ -18,6 +25,13 @@ class QuickstartUser(HttpUser):
     @tag('sync_classification')
     @task
     def sync_classification(self):
-        img_path = '/home/lk/MyProject/BMW_F3OCR/数据集/文件分类/营业执照/授信资料-43.jpg'
+        img_path = random.choice(file_path_list) 
         files=[('image', ('', open(img_path,'rb'), ''))]
         self.client.post("/sync_classification", files=files)
+
+    @tag('async_classification')
+    @task
+    def async_classification(self):
+        img_path = random.choice(file_path_list)
+        files=[('image', ('', open(img_path,'rb'), ''))]
+        self.client.post("/async_classification", files=files)

@@ -29,15 +29,68 @@ tf_serving_settings = {
 app.config.update(tf_serving_settings)
 
 
-@app.post("/sync_classification")
-async def sync_handler(request):
+# 同步写法01
+# @app.post("/sync_classification")
+# async def sync_handler(request):
+#     image = request.files.get("image")
+#     img_array = np.frombuffer(image.body, np.uint8)
+#     image = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
+#     input_images = classifier.preprocess_input(image)
+# 
+#     options = [('grpc.max_send_message_length', 1000 * 1024 * 1024),
+#                ('grpc.max_receive_message_length', 1000 * 1024 * 1024)]
+#     with grpc.insecure_channel('localhost:8500', options=options) as channel:
+#         stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
+#         # See prediction_service.proto for gRPC request/response details.
+#         request = predict_pb2.PredictRequest()
+#         request.model_spec.name = classifier.model_name
+#         request.model_spec.signature_name = classifier.signature_name
+# 
+#         request.inputs['input_1'].CopyFrom(tf.make_tensor_proto(input_images))
+#         result = stub.Predict(request, timeout=100.0)  # 100 secs timeout
+#         outputs = tf.make_ndarray(result.outputs['output'])
+# 
+#     res = classifier.reprocess_output(outputs)
+#     return json(res)
+
+# 同步写法02
+# @app.post("/sync_classification")
+# async def sync_handler(request):
+#     image = request.files.get("image")
+#     img_array = np.frombuffer(image.body, np.uint8)
+#     image = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
+#     input_images = classifier.preprocess_input(image)
+# 
+#     # See prediction_service.proto for gRPC request/response details.
+#     request = predict_pb2.PredictRequest()
+#     request.model_spec.name = classifier.model_name
+#     request.model_spec.signature_name = classifier.signature_name
+#     stub = getattr(app, classifier.server_name)
+# 
+#     request.inputs['input_1'].CopyFrom(tf.make_tensor_proto(input_images))
+#     result = stub.Predict(request, timeout=100.0)  # 100 secs timeout
+#     outputs = tf.make_ndarray(result.outputs['output'])
+# 
+#     res = classifier.reprocess_output(outputs)
+#     return json(res)
+# 
+# @app.listener("before_server_start")
+# async def set_grpc_channel(app, loop):
+#     for server_name, server_settings in app.config['servers'].items():
+#         channel = grpc.insecure_channel(
+#             '{0}:{1}'.format(server_settings['host'], server_settings['port']),
+#             options=server_settings.get('options'))
+#         stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
+#         setattr(app, server_name, stub)
+
+# 异步写法
+@app.post("/async_classification")
+async def async_handler(request):
     image = request.files.get("image")
     img_array = np.frombuffer(image.body, np.uint8)
     image = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
     input_images = classifier.preprocess_input(image)
 
-    # print(type(image))
-
     # See prediction_service.proto for gRPC request/response details.
     request = predict_pb2.PredictRequest()
     request.model_spec.name = classifier.model_name
@@ -45,23 +98,17 @@ async def sync_handler(request):
     stub = getattr(app, classifier.server_name)
 
     request.inputs['input_1'].CopyFrom(tf.make_tensor_proto(input_images))
-    result = stub.Predict(request, 100.0)  # 100 secs timeout
+    result = await stub.Predict(request, timeout=100.0)  # 100 secs timeout
     outputs = tf.make_ndarray(result.outputs['output'])
 
     res = classifier.reprocess_output(outputs)
     return json(res)
 
 
-# @app.get("/async")
-# async def async_handler(request):
-#     await asyncio.sleep(2)
-#     return json({'code': 1})
-
-
 @app.listener("before_server_start")
 async def set_grpc_channel(app, loop):
     for server_name, server_settings in app.config['servers'].items():
-        channel = grpc.insecure_channel(
+        channel = grpc.aio.insecure_channel(
             '{0}:{1}'.format(server_settings['host'], server_settings['port']),
             options=server_settings.get('options'))
         stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
@@ -69,4 +116,4 @@ async def set_grpc_channel(app, loop):
 
 
 if __name__ == '__main__':
-    app.run(host='0.0.0.0', port=6699, workers=5)
+    app.run(host='0.0.0.0', port=6699, workers=10)