# jobot_factory_simple_nlp

**Repository Path**: jeusgao/jobot_factory_simple_nlp

## Basic Information

- **Project Name**: jobot_factory_simple_nlp
- **Description**: TF Model building/training/evaluating for simple nlp task just by params configuration, training/evaluating monitor and params configure GUI with streamlit.
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 5
- **Forks**: 0
- **Created**: 2021-01-23
- **Last Updated**: 2024-03-03

## Categories & Tags

**Categories**: machine-learning

**Tags**: None

## README

A light tool for TF Model building/training/evaluating for simple nlp tasks just by params configuration, training/evaluating monitor and params configure GUI with streamlit, and auto publish all models' predictor apis

copy pretrained LM into hub/bases/ (bert, rbt, albert ...)

## How to run:
1. run init_params first to generate params template...
2. streamlit run home.py --server.port=PORT
3. celery -A celery_training worker -l INFO --pidfile=celery/%n.pid --logfile=celery/%n%I.log

open localhost:PORT with a broswer ...

## Rest service:
* uvicorn rest_service.handlers:app --port=REST_SERVER-PORT or python3 rest_server.py -p=REST_SERVER-PORT

## ENV
python >= 3.6
</br>
test with tf 2.4.0, 2.2.2
</br>

## Preview
### *common params setting*
<img src="imgs/4.jpg" />

### *model layers settings*
<img src="imgs/3.jpg" />

### *model layers' params settings*
<img src="imgs/2.jpg" />

### *trainning monitoring*
<img src="imgs/6.jpg" />

### *trainning scores*
<img src="imgs/5.jpg" />

### *api published automaticly*
<img src="imgs/1.jpg" />

## references:
#### <a href="https://github.com/CyberZHG/keras-bert">keras-bert</a>
#### <a href="https://github.com/BrikerMan/Kashgari">Kashgari</a>
#### <a href="https://github.com/bojone/bert4keras">Bert4Keras</a>