Bert-Server

Dockerfiles for building based Bert-server images.

Supported tags and respective Dockerfile links

phenompeople/bert-server

• latest (UNCASED_EN_BASE/Dockerfile)
• UNCASED_EN_BASE (UNCASED_EN_BASE/Dockerfile)
• UNCASED_EN_LG (UNCASED_ENG_LG/Dockerfile)
• UNCASED_EN_LG-GPU (UNCASED_EN_LG-GPU/Dockerfile)

Pre-Requisites

• install docker-engine https://docs.docker.com/engine/installation/

Whats New?

Using BERT model as a sentence encoding service, i.e. mapping a variable-length sentence to a fixed-length vector.

How to use this image

1. This image can be used by simply running

$ docker run --name=bert-server -p 5555:5555 -p 5556:5556 -td phenompeople/bert-server

Above command runs bert-server container with port 5555and 5556 mapped to host and connecting to bert meodles running.

1. To make image run even after reboot use extra option --restart=always

$ docker run --restart=always --name=bert-server -p 5555:5555 -p 5556:5556 -td phenompeople/bert-server

Maintainers

• Gaurav Vijay Vergia (gaurav.vergiya@phenompeople.com)
• Rajesh Jonnalagadda (rajesh.jonnalagadda@phenompeople.com)

License and Authors

License: Apache License

Author : Phenompeople Pvt Ltd (devops@phenompeople.com)