https://github.com/krystianity/keras-serving

bring keras-models to production with tensorflow-serving and nodejs + docker :pizza:
https://github.com/krystianity/keras-serving

cpp docker grpc keras network neuronal nodejs production python serving tensorflow

Last synced: 9 days ago
JSON representation

bring keras-models to production with tensorflow-serving and nodejs + docker :pizza:

• Host: GitHub
• URL: https://github.com/krystianity/keras-serving
• Owner: krystianity
• License: apache-2.0
• Created: 2017-05-19T22:58:56.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2022-12-09T16:54:32.000Z (almost 2 years ago)
• Last Synced: 2024-04-15T07:39:12.379Z (7 months ago)
• Topics: cpp, docker, grpc, keras, network, neuronal, nodejs, production, python, serving, tensorflow
• Language: Python
• Homepage:
• Size: 20.7 MB
• Stars: 153
• Watchers: 16
• Forks: 27
• Open Issues: 13
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# keras-serving

keras->tensorflow+grpc+docker=>nodejs :whale::fire:

- example of bringing a keras model to production using tensorflow serving

- using custom XOR model with tensor.proto dimensions example

- building & training of the model works with python2.7 on the workstation

- exported model is served via grpc in a C++ server using a Docker-Container

- a nodejs server wraps the grpc api for a simple http POST endpoint

- also ships an advanced multi-model face (emotion, gender) detection example `/face-recog`

- and and advanced google bigquery (as dataaset) example `/bigquery`

# Overview

* XOR Setup below

* [Face-Recog Setup](face-recog/)

* [BigQuery Setup](bigquery/)

# Workflow (Unix - testen on Ubuntu 16.04 64bit)

## You will need

```

python

pip

docker (docker-compose)

```

## 1. Install Requirements

```

./prepare.sh

# installs python dependencies via pip

# builds the docker image for tensorflow_serving (takes a while ~ 30 minutes)

# image size ~ 3.5 GB

```

checkout [build troubleshoot](build.md) if you are having trouble

## 2. Build, Train and Serialise Keras Model

```

python train.py

# results will be in (/result)

```

## 3. Load and export Model as Tensorflow Graph

```

python export.py

# results will be in (/export)

```

## 4. Build & Run Containers via docker-compose

```

./start-servers.sh

# ./stop-servers.sh

```

## 5. Test API via curl

```

curl -X POST \

  http://localhost:8080/predict-xor \

  -H 'cache-control: no-cache' \

  -H 'content-type: application/json' \

  -d '{"inputs": [0,1]}'

```