hboard-widget

hboard-widget

中文

This project provides a visualization tool for experiment information based on jupyter notebook/ jupyterlab widget.

Installation

Install with pip

pip install hboard-widget

Install with conda

conda install -c conda-forge hboard-widget

Install with source code

Build from source code need following requirements:

• python 3.7+
• nodejs v14.15.0+
• pip 20.0.2+
• jupyterlab 2.0.0+

The project need frontend of hboard to be built in advance.

Clone the source code:

git clone https://github.com/DataCanvasIO/HyperBoard.git

Build for development:

cd ./hboard-widget
pip install -e .
jupyter nbextension install --py --symlink --overwrite --sys-prefix hboard_widget
jupyter nbextension enable --py --sys-prefix hboard_widget
jupyter labextension develop --overwrite hboard_widget

You need to rebuild the JS when you make a code change:

cd ./hboard-widget/js
yarn run build

Example

The following steps shows how to implement the experiment visualization in notebook

1. Import the required modules：

import warnings
warnings.filterwarnings('ignore')

from sklearn.model_selection import train_test_split
from hypernets.examples.plain_model import PlainModel, PlainSearchSpace
from hypernets.experiment import make_experiment
from hypernets.tabular.datasets import dsutils

2. Create an experiment

df = dsutils.load_boston()
df_train, df_eval = train_test_split(df, test_size=0.2)
search_space = PlainSearchSpace(enable_lr=False, enable_nn=False, enable_dt=False, enable_dtr=True)
experiment = make_experiment(PlainModel, df_train,
                             target='target',
                             search_space=search_space,
                             callbacks=[],
                             search_callbacks=[])

3. Experiment visualization configurations

from hboard_widget import ExperimentSummary
experiment_summary_widget = ExperimentSummary(experiment)
display(experiment_summary_widget)

4. Visualize the dataset information

from hboard_widget import DatasetSummary
dataset_summary_widget = DatasetSummary(experiment.get_data_character())
display(dataset_summary_widget)

5. Visualize the experiment process

from hboard_widget import ExperimentProcessWidget
estimator = experiment.run(max_trials=3)

widget = ExperimentProcessWidget(experiment)
display(widget)

Find the project in Notebook experiment visualization notebook.ipynb.

Related project

Currently, HyperGBM has integrated this tool. The HyperGBM experiment could call the notebook widget visualization function and display the experiment dashboard. Please refer to HyperGBM: Experiment Visualization in Notebook