Command Line Interface

The CLI allows OpenKiwi to carry it’s pipelines without having to import kiwi as a python module. After calling one of these pipelines, the specific parameters can be configured by using the flags available in the configuration options

Quality Estimation toolkit

usage: kiwi [-h] [--version] {train,predict,jackknife,evaluate} ...

Named Arguments

--version show program’s version number and exit

Pipelines

Use ‘kiwi <pipeline> (-h | –help)’ to check it out.

pipeline

Possible choices: train, predict, jackknife, evaluate

Available pipelines:

Sub-commands:

train

Train a QE model

For train configuration options see Train Interface.

kiwi train

predict

Use a pre-trained model for prediction

For prediction configuration options see Predict Interface.

kiwi predict

jackknife

Jackknife training data with model

For jackknife configuration options see jackknife Interface.

kiwi jackknife

evaluate

Evaluate a model’s predictions using popular metrics

For evaluation configuration options see Evaluate Interface.

kiwi evaluate

Copyright 2019 Unbabel. All rights reserved. Source code available under the AGPL-3.0.