twinlab 1.2.0

Client interface for twinLab machine-learning in the cloud.

twinLab Client

Headless interface to the twinLab library.

Installation

Most users should use pip

pip install twinlab

If you want to modify the client-side code, or have a local installation, you will need to have git, poetry, and a python version of 3.9 or higher installed. Then you can do:

git clone https://github.com/digiLab-ai/twinLab-client.git
cd twinlab-client
poetry install

Environment setup

You will need a .env file in your project directory that looks like the .env.example file in this repository

cp .env.example .env

and fill in your twinLab user details.

Commands

Pipeline

poetry run python scripts/twinlab/test.py

where test.py can be replaced with any of the scripts in the scripts directory.

Individual examples

Get user information:

poetry run python scripts/twinlab/get_user.py

Get version information:

poetry run python scripts/twinlab/get_versions.py

List datasets:

poetry run python scripts/twinlab/list_datasets.py

Upload dataset:

poetry run python scripts/twinlab/upload_dataset.py <path/to/dataset.csv> <dataset_id>
poetry run python scripts/twinlab/upload_dataset.py resources/datasets/biscuits.csv biscuits

View dataset:

poetry run python scripts/twinlab/view_dataset.py <dataset_id>
poetry run python scripts/twinlab/view_dataset.py biscuits

Summarise a dataset:

poetry run python scripts/twinlab/summarise_dataset.py <dataset_id>
poetry run python scripts/twinlab/summarise_dataset.py biscuits

List campaigns:

poetry run python scripts/twinlab/list_campaigns.py

Train campaign:

poetry run python scripts/twinlab/train_campaign.py <path/to/parameters.json> <campaign_id> <processor>
poetry run python scripts/twinlab/train_campaign.py resources/campaigns/biscuits/params.json biscuits-campaign

Get campaign status:

poetry run python scripts/twinlab/status_campaign.py <campaign_id>
poetry run python scripts/twinlab/status_campaign.py biscuits-campaign

Summarise campaign:

poetry run python scripts/twinlab/summarise_campaign.py <campaign_id>
poetry run python scripts/twinlab/summarise_campaign.py biscuits-campaign

Predict using a trained campaign:

poetry run python scripts/twinlab/predict_campaign.py <path/to/inputs.csv> <campaign_id> <method> <processor>
poetry run python scripts/twinlab/predict_campaign.py resources/campaigns/biscuits/eval.csv biscuits-campaign

Delete a campaign:

poetry run python scripts/twinlab/delete_campaign.py <campaign_id>
poetry run python scripts/twinlab/delete_campaign.py biscuits-campaign

Delete a dataset:

poetry run python scripts/twinlab/delete_dataset.py <dataset_id>
poetry run python scripts/twinlab/delete_dataset.py biscuits

Full example

Here we create some mock data (which has a quadratic relationship between X and y) and use twinLab to create a surrogate model with quantified uncertainty.

# Import libraries
import twinlab as tl
import pandas as pd

# Create a dataset and upload to twinLab cloud
df = pd.DataFrame({"X": [1, 2, 3, 4], "y": [1, 4, 9, 16]})
tl.upload_dataset(df, "test-data")

# Train a machine-learning model for the data
params = {
    "dataset_id": "test-data",
    "inputs": ["X"],
    "outputs": ["y"],
}
tl.train_campaign(params, campaign_id="test-model")

# Evaluate the model on some unseen data
df = pd.DataFrame({"X": [1.5, 2.5, 3.5]})
df_mean, df_std = tl.predict_campaign(df, campaign_id="test-model")

Notebooks

Check out the notebooks directory for some additional examples to get started!

Documentation

See the live documentation at https://digilab-ai.github.io/twinLab-client/. Or build a copy locally:

cd docs
yarn install && yarn start