提交 21032ebc 编写于 作者: J Jiawei Wang 提交者: GitHub

Update BERT_10_MINS.md

上级 6561b36c
...@@ -8,6 +8,7 @@ The goal of Bert-As-Service is to give a sentence, and the service can represent ...@@ -8,6 +8,7 @@ The goal of Bert-As-Service is to give a sentence, and the service can represent
Paddle Serving supports various models trained based on Paddle, and saves the serviceable model by specifying the input and output variables of the model. For convenience, we can load a trained bert Chinese model from paddlehub and save a deployable service with two lines of code. The server and client configurations are placed in the `bert_seq20_model` and` bert_seq20_client` folders, respectively. Paddle Serving supports various models trained based on Paddle, and saves the serviceable model by specifying the input and output variables of the model. For convenience, we can load a trained bert Chinese model from paddlehub and save a deployable service with two lines of code. The server and client configurations are placed in the `bert_seq20_model` and` bert_seq20_client` folders, respectively.
[//file]:#bert_10.py
``` python ``` python
import paddlehub as hub import paddlehub as hub
model_name = "bert_chinese_L-12_H-768_A-12" model_name = "bert_chinese_L-12_H-768_A-12"
...@@ -27,6 +28,7 @@ serving_io.save_model("bert_seq20_model", "bert_seq20_client", ...@@ -27,6 +28,7 @@ serving_io.save_model("bert_seq20_model", "bert_seq20_client",
#### Step2: Launch Service #### Step2: Launch Service
[//file]:#server.sh
``` shell ``` shell
python -m paddle_serving_server_gpu.serve --model bert_seq20_model --thread 10 --port 9292 --gpu_ids 0 python -m paddle_serving_server_gpu.serve --model bert_seq20_model --thread 10 --port 9292 --gpu_ids 0
``` ```
...@@ -43,6 +45,7 @@ Paddle Serving has many built-in corresponding data preprocessing logics. For th ...@@ -43,6 +45,7 @@ Paddle Serving has many built-in corresponding data preprocessing logics. For th
Install paddle_serving_app Install paddle_serving_app
[//file]:#pip_app.sh
```shell ```shell
pip install paddle_serving_app pip install paddle_serving_app
``` ```
...@@ -51,6 +54,7 @@ pip install paddle_serving_app ...@@ -51,6 +54,7 @@ pip install paddle_serving_app
the script of client side bert_client.py is as follow: the script of client side bert_client.py is as follow:
[//file]:#bert_client.py
``` python ``` python
import os import os
import sys import sys
...@@ -71,6 +75,7 @@ for line in sys.stdin: ...@@ -71,6 +75,7 @@ for line in sys.stdin:
run run
[//file]:#bert_10_cli.sh
```shell ```shell
cat data.txt | python bert_client.py cat data.txt | python bert_client.py
``` ```
...@@ -82,3 +87,18 @@ read samples from data.txt, print results at the standard output. ...@@ -82,3 +87,18 @@ read samples from data.txt, print results at the standard output.
We tested the performance of Bert-As-Service based on Padde Serving based on V100 and compared it with the Bert-As-Service based on Tensorflow. From the perspective of user configuration, we used the same batch size and concurrent number for stress testing. The overall throughput performance data obtained under 4 V100s is as follows. We tested the performance of Bert-As-Service based on Padde Serving based on V100 and compared it with the Bert-As-Service based on Tensorflow. From the perspective of user configuration, we used the same batch size and concurrent number for stress testing. The overall throughput performance data obtained under 4 V100s is as follows.
![4v100_bert_as_service_benchmark](4v100_bert_as_service_benchmark.png) ![4v100_bert_as_service_benchmark](4v100_bert_as_service_benchmark.png)
<!--
yum install -y libXext libSM libXrender
pip install paddlehub paddle_serving_server paddle_serving_client
sh pip_app.sh
python bert_10.py
sh server.sh &
wget https://paddle-serving.bj.bcebos.com/bert_example/data-c.txt --no-check-certificate
cat data-c.txt | python bert_client.py
if [[ $? -eq 0 ]]; then
echo "test success"
else
echo "test fail"
fi
-->
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册