Sign inSign up

nishizhen/opus-mt

By nishizhen

Updated over 1 year ago

This image is build without Intel MKL, Upgrade Marian to 1.12.0, Add test pair eng-zho.

Image
Languages & frameworks
Machine learning & AI
Data science
0

459

nishizhen/opus-mt repository overview

This image is build without Intel MKL, Upgrade Marian to 1.12.0, Add test pair eng-zho.

version: '3'

services:
  opus-mt:
    image: nishizhen/opus-mt:latest
    volumes:
      - /models:/usr/src/app/models
      - /confit/services.json:/usr/src/app/services.json
    ports:
      - 8888:80

services.json example:

{
	"en": {
		"es": {
			"configuration": "./models/en-es/decoder.yml",
			"sourcebpe": "./models/en-es/source.bpe",
			"host": "localhost",
			"port": "10001"
		},
		"fi": {
			"configuration": "./models/en-fi/decoder.yml",
			"sourcespm": "./models/en-fi/source.spm",
			"targetspm": "./models/en-fi/target.spm",
			"host": "localhost",
			"port": "10002"
		},
		"zh": {
			"configuration": "./models/eng-zho/decoder.yml",
			"sourcespm": "./models/eng-zho/source.spm",
			"targetspm": "./models/eng-zho/target.spm",
			"host": "localhost",
			"port": "10003"
		}
	}
}

Download the translation models from https://github.com/Helsinki-NLP/OPUS-MT-train/tree/master/models ,https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models and place them in models directory.

Then edit the services.json to point to those models.

Intro of Opus-MT :https://github.com/Helsinki-NLP/Opus-MT/

Tag summary

Content type

Image

Digest

sha256:b4b54d8af

Size

181.4 MB

Last updated

over 1 year ago

docker pull nishizhen/opus-mt