zero2deploy's Introduction

zero2Deploy

Sentiment Analysis based on Tensorflow2.0 Kears with deploy

Requirements

python3
tensorflow >= 2.0

data & preprocess

The data provided in the data/ directory is a csv file

In data_util.py I provide some funtions to process the csv file.

Usage

This contains several steps:

Before you can get started on training the model, you mast run

python data_util.py

After the dirty preprpcessing jobs, you can try running an training experiment with some configurations by:

python train.py

After that, you get a fold named "my_cls_model", you can copy the fold to the deploy machine. And run:

# install docker
sudo apt-get install docker # ignore if you have already install docker
# deploy serveing 
docker pull tensorflow/serving
# ignore steps above if you have already installed docker
docker run -it --rm -p 8500:8500 -p 8501:8501 -v "/my_cls_model:/models/my_cls_model" -e MODEL_NAME=my_cls_model tensorflow/serving

After the server is done, the you can run:

python server_test.py

to test the server connection is good.

Then you need start the Flask by:

python app.py

Finally you can send the request to serving by:

python simple_request.py

Follow the instruction. Hope you enjoy it.

Reference

Using Tensorflow2.0 Keras to deply sentiment analysis model from scratch .

Folder Structure

├── data            - this fold contains all the data
│   ├── train
│   ├── dev
│   ├── test
│   ├── vocab
|   ├── vec
├── my_cls_model           - this fold contains the pb file to restore
├── train.py               - main entrance of the project
├── data_util.py           - preprocess the data
├── load_data.py           - data generator
├── server_test.py       - test if net works after deploy 
├── app.py                 - use flask  
├── simple_request.py      - use rest to send query