An open source project from Data to AI Lab at MIT.

D3M Dataset Manager

The D3M Dataset Manager is a command line tool and python package to generate and manage datasets in the D3M format.

• Documentation: https://HDI-Project.github.io/d3m-dataset-manager
• Homepage: https://github.com/HDI-Project/d3m-dataset-manager

Overview

The D3M Dataset Manager is a command line tool and python package to generate and manage datasets in the D3M format.

It supports:

• downloading datasets from the D3M web repository or from S3 buckets
• uploading datasets to S3 buckets
• loading or saving datasets to local filesystem
• spliting datasets into TRAIN, TEST and SCORE subsets following the dataSplits.csv indexes

Data Format

The D3M Dataset Schema, developed by MIT Lincoln Labs Laboratory for the DARPA's Data Driven Discovery of Models Program, requires the data to be in plainly readable formats such as CSV files or JPG images, and to be set within a folder hierarchy alongside some metadata specifications in JSON format, which include information about all the data contained, as well as the problem that we are trying to solve.

For more details about the schema and about how to format your data to be compliant with it, please have a look at the Schema Documentation

Install

Install from PyPI

The easiest and recommended way to install the D3M Dataset Manager is using pip:

pip install d3m-dataset-manager

This will pull and install the latest stable release from PyPI.

Install from source

If you want to install the project from its sources, you can clone the repository and install it by running make install on the stable branch:

git clone [email protected]:HDI-Project/d3m-dataset-manager.git
cd d3m-dataset-manager
git checkout stable
make install

Install for Development

If you want to contribute to the project, a few more steps are required to make the project ready for development.

Please head to the Contributing Guide for more details about this process.

Usage

Configuration

D3M Repository

In order to interact with the D3M repository you will need the user and the password user to log into https://datadrivendiscovery.org/data

S3 Bucket

In order to interact with the S3 buckets, you will need to configure your S3 access following the instructions from http://boto3.readthedocs.io/en/latest/guide/quickstart.html

In most cases, it will be enough to create the file ~/.aws/credentials: with the following contents:

[default]
aws_access_key_id = YOUR_ACCESS_KEY
aws_secret_access_key = YOUR_SECRET_KEY

Command Line Options

The main element of the D3M Dataset Manager is the commadn d3mdm, which will be available in your command line after installing the package.

This command supports the following options:

• -i, --input - D3M website, IPFS, S3 bucket or local folder.
• -o, --output - S3 bucket or local folder.
• -l, --list - List all available datasets in the indicated input.
• -a, --all - Get and process all available datasets in the indicated input.
• -s, --split - Split the dataset using the dataSplits.csv indexes.
• -r, --raw - Do not download the splitted subsets. -s option implicitly enables this one.
• -f, --force - Overwrite any existing datasets. If not enabled, existing datasets will be skipped.
• -d, --dry-run - Do not perform any real action. Only list them.
• dataset names - Name of the datasets o download. The -a option overrides them.

Input and Output

The Input and Output options implicitely point at different locations depending on the format:

• D3M: d3m:username:passsword: password can be omitted, as well as username. Accepted only as Input. If omitted, the user will be asked to insert them later on.
• IPFS: ipfs: The datasets will be downloaded using an IPFS mirror of the D3M repository.
• S3: s3://bucket-name/folder: The datasets will be stored as a .tar.gz archive. If folder is not specified it defaults to datasets.
• Local filesystem: local/filesystem/path: The path must exist, otherwise it raises an error.

Usage Examples

Download all datasets from D3M and store them as they are into S3 bucket named d3m-data-dai. This will skip existing datasets.

d3m-dataset-manager -i d3m:a_username:a_password -o s3:d3m-data-dai -a

Download all datasets from the IPFS mirror, split them and store them in a local folder datasets, overwriting any existing data.

This will prompt the user for the d3m password.

d3m-dataset-manager -i ipfs -o datasets -a -s -f

Download the datasets 185_baseball and 32_wikiqa from S3 bucket bucket-name into local folder data/datasets. Overwrite the existing data.

d3m-dataset-manager -i s3://bucket-name -o data/datasets -f 185_baseball 32_wikiqa

What's next?

For more details about D3M Dataset Manager and all its possibilities and features, please check the documentation site.