def multithread_http(thread, batch_size):
multi_thread_runner = MultiThreadRunner()
start = time.time()
result = multi_thread_runner.run(run_http, thread, batch_size)
end = time.time()
total_cost = end - start
avg_cost = 0
total_number = 0
for i in range(thread):
avg_cost += result[0][i]
total_number += result[2][i]
avg_cost = avg_cost / thread
print("Total cost: {}s".format(total_cost))
print("Each thread cost: {}s. ".format(avg_cost))
print("Total count: {}. ".format(total_number))
print("AVG QPS: {} samples/s".format(batch_size * total_number /
total_cost))
show_latency(result[1])
else:
raise ValueError("not implemented {} request".format(args.request))
end = time.time()
if latency_flags:
return [[end - start], latency_list, [total_number]]
else:
return [[end - start]]
if __name__ == '__main__':
multi_thread_runner = MultiThreadRunner()
endpoint_list = ["127.0.0.1:9393"]
turns = 1
start = time.time()
result = multi_thread_runner.run(single_func, args.thread, {
"endpoint": endpoint_list,
"turns": turns
})
end = time.time()
total_cost = end - start
total_number = 0
avg_cost = 0
for i in range(args.thread):
avg_cost += result[0][i]
total_number += result[2][i]
avg_cost = avg_cost / args.thread
print("total cost-include init: {}s".format(total_cost))
print("each thread cost: {}s. ".format(avg_cost))
print("qps: {}samples/s".format(args.batch_size * total_number /
(avg_cost * args.thread)))
print("qps(request): {}samples/s".format(total_number /
return [[end - start]]
if __name__ == '__main__':
"""
usage:
"""
img_file_list = get_img_names("./img_data")
img_content_list = preprocess_img(img_file_list)
multi_thread_runner = MultiThreadRunner()
endpoint_list = ["127.0.0.1:9494"]
turns = 1
start = time.time()
result = multi_thread_runner.run(benckmark_worker, args.thread, {
"endpoint": endpoint_list,
"turns": turns,
"img_list": img_content_list
})
end = time.time()
total_cost = end - start
avg_cost = 0
for i in range(args.thread):
avg_cost += result[0][i]
avg_cost = avg_cost / args.thread
print("total cost: {}s".format(total_cost))
print("each thread cost: {}s. ".format(avg_cost))
print("qps: {}samples/s".format(args.batch_size * args.thread * turns /
total_cost))
if os.getenv("FLAGS_serving_latency"):
show_latency(result[1])
def multithread_rpc(thraed, batch_size):
multi_thread_runner = MultiThreadRunner()
result = multi_thread_runner.run(run_rpc, thread, batch_size)
# limitations under the License.
from paddle_serving_client import Client
from paddle_serving_client.utils import MultiThreadRunner
import paddle
import numpy as np
def single_func(idx, resource):
client = Client()
client.load_client_config(
"./uci_housing_client/serving_client_conf.prototxt")
client.connect(["127.0.0.1:9293", "127.0.0.1:9292"])
x = [
0.0137, -0.1136, 0.2553, -0.0692, 0.0582, -0.0727, -0.1583, -0.0584,
0.6283, 0.4919, 0.1856, 0.0795, -0.0332
]
x = np.array(x)
for i in range(1000):
fetch_map = client.predict(feed={"x": x}, fetch=["price"])
if fetch_map is None:
return [[None]]
return [[0]]
multi_thread_runner = MultiThreadRunner()
thread_num = 4
result = multi_thread_runner.run(single_func, thread_num, {})
if None in result[0]:
exit(1)
def single_func(idx, resource):
if args.request == "rpc":
client = Client()
client.load_client_config(args.model)
client.connect([args.endpoint])
train_reader = paddle.batch(paddle.reader.shuffle(
paddle.dataset.uci_housing.train(), buf_size=500),
batch_size=1)
start = time.time()
for data in train_reader():
fetch_map = client.predict(feed={"x": data[0][0]}, fetch=["price"])
end = time.time()
return [[end - start]]
elif args.request == "http":
train_reader = paddle.batch(paddle.reader.shuffle(
paddle.dataset.uci_housing.train(), buf_size=500),
batch_size=1)
start = time.time()
for data in train_reader():
r = requests.post('http://{}/uci/prediction'.format(args.endpoint),
data={"x": data[0]})
end = time.time()
return [[end - start]]
multi_thread_runner = MultiThreadRunner()
result = multi_thread_runner.run(single_func, args.thread, {})
print(result)
def multithread_http(thread, batch_size):
multi_thread_runner = MultiThreadRunner()
result = multi_thread_runner.run(run_http, thread, batch_size)
def multithread_http(args):
multi_thread_runner = MultiThreadRunner()
result = multi_thread_runner.run(do_client, args.threads, args)