Metadata-Version: 2.1
Name: bert-multitask-server
Version: 0.1.1
Summary: A service to serve bert_multitask_learning models(server)
Home-page: https://github.com/JayYip/bert-multitask-service
Author: Jay Yip
Author-email: junpang.yip@gmail.com
License: MIT
Keywords: bert nlp tensorflow machine learning sentence encoding embedding serving
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3.6
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Description-Content-Type: text/markdown
Requires-Dist: setuptools (==40.5.0)
Requires-Dist: tqdm (==4.28.1)
Requires-Dist: six (==1.11.0)
Requires-Dist: tensorflow-gpu (>=1.12.0)
Requires-Dist: numpy (==1.15.4)
Requires-Dist: tensor2tensor (==1.11.0)
Requires-Dist: termcolor (==1.1.0)
Requires-Dist: GPUtil (==1.4.0)
Requires-Dist: protobuf (==3.6.1)
Requires-Dist: scikit-learn (==0.20.2)
Requires-Dist: pyzmq (==17.1.2)


# bert-multitask-as-service

A serving service for [bert-multitask-learning](https://github.com/JayYip/bert-multitask-learning)

## Install

```bash
pip install bert-multitask-server
pip install bert-multitask-client
```

## Getting Started

1. Train and export model.

    A typical trained checkpoint dir looks like below.

    ```text
    bert_serving_ckpt/
    ├── *_label_encoder.pkl
    ├── bert_config.json
    ├── export_model
    ├── params.json
    └── vocab.txt
    ```

2. Start server using CLI

    ```bash
    bert-multitask-serving-start -model_dir ~/CWS_NER_POS_ckpt/ -num_worker=4 -problem "CWS|NER|POS"
    ```

3. Use Client to Get Prediction

    ```python
    from bert_serving.client import BertClient
    bc = BertClient()
    bc.encode(['我爱北京天安门'])
    ```

# Bert多任务学习服务

一个部署[Bert多任务学习](https://github.com/JayYip/bert-multitask-learning)的服务

## 安装

```bash
pip install bert-multitask-server
pip install bert-multitask-client
```

## 开始使用

1. 训练模型并导出模型.

    导出后的模型目录应该有以下文件

    ```text
    bert_serving_ckpt/
    ├── *_label_encoder.pkl
    ├── bert_config.json
    ├── export_model
    ├── params.json
    └── vocab.txt
    ```

2. 用CLI启动服务

    ```bash
    bert-multitask-serving-start -model_dir models/ -num_worker=4 -problem "fake_problem"
    ```

3. 用客户端获取预测结果

    ```python
    from bert_serving.client import BertClient
    bc = BertClient()
    bc.encode(['我爱北京天安门'])
    ```


