Usage

Install

This package is published as a Gitlab PyPI package. To install it, you need to:

  1. Create a personal access token with scope read_api and access to this repository.

  2. Add the following to your ~/.netrc file:

     machine gitlab.teklia.com
 login __token__
 password <YOUR_PERSONAL_TOKEN>

For development and tests purpose it may be useful to install the worker as an editable package with pip.

# The `--index-url` argument is required to find the teklia-qwen package.
pip3 install --index-url https://gitlab.teklia.com/api/v4/projects/403/packages/pypi/simple -e .

Commands

When teklia-qwen is installed in your environment, you may use the following commands:

teklia-qwen dataset

To generate a training dataset from an Arkindex SQLite export. More details in the dedicated page.

teklia-qwen train

To fine-tune a model using Low Rank Adaptation (LoRA). More details in the dedicated page.

teklia-qwen predict

To predict on multiple images. More details in the dedicated page.

teklia-qwen evaluate

To evaluate predictions. More details in the dedicated page.

teklia-qwen convert

To convert predictions to same format used in DAN. More details in the dedicated page.

teklia-qwen merge

To merge the adapter layers into the base model. More details in the dedicated page.

Scripts

Helper scripts are available in this page.