mlflow-ngl

Background

The project was partially forked from MLflow dashboard, which I added some customized React Component with NGL for chemoinformatics visualization.

Follow the style of the MLflow artifacts, all the customized components would be located under this path: src/experiment-tracking/components/artifact-nglview-components.

Setup

1. Install Python Dependencies

There are several Python Dependencies which would be used:

pyyaml
flask
mlflow

I would recommend you to create a conda environment to install all the Python dependencies above, see more details about the Miniconda installation.

After install the Miniconda, you can create a conda environment by:

conda create -n mlflow-ngl python=3.7

Then install the dependencies by:

conda install -c conda-forge pyyaml flask mlflow

1. Modify artifacts path in the meta.yaml

Active the Python/Conda environment you just setup in step 1. and run the command below to modify the absolute file path in the meta.yaml.

make artifacts

1. Build/install npm packages

yarn

1. Run server

make dev

Development

A sample of NGLView example is showed in src/experiment-tracking/components/artifact-nglview-components. If you want to add a customized view of the NGL component, please follow the steps below:

1. Add a new extension support in ChemFileUtils.js, for example:

export const STRUCTURE_EXTENSIONS = new Set(['pdb', 'sdf']);

1. Modify the render() of ShowNGLPage to add a new extension, for example:

if (this.props.path) {
   const normalizedExtension = getExtension(this.props.path);
   if (normalizedExtension) {
       if (STRUCTURE_EXTENSIONS.has(normalizedExtension.toLowerCase())) {
           return <ShowNGLStructureView runUuid={this.props.runUuid} path={this.props.path} ext={normalizedExtension} />;
       }
   }
}

1. Prepare the testing data

Rename the testing file with specific extension and move it to one ./mlruns/ experiments for testing.

1. Rerun the app through make dev

trumanw/mlflow-ngl

mlflow-ngl

Background

Setup

Development