In this section, we describe how to prepare and share your run with the RanxHub community!
We encourage you to consider sharing your run to help others save time, resources, and reduce CO₂ emissions.

What is a Run Card?

A run card is a YAML file containing your run's metadata. The metadata that you add to the run card improve its discoverability and enable certain interactions on the hub, such as indexing and tag-based filtering.

A run card is divided into four sections:

• Run metadata: how the run was computed, who computed the run, tags for indexing, and other information;
• Benchmark metadata: name, dataset, split, version, and the ir-datasets id, if any;
• Paper metadata: title, authors (with name and ORCID), publication date, doi, dblp URL for BibTex;
• Model metadata: name, description, tags for indexing, paper title, authors (with name and ORCID), publication date, doi, dblp URL for BibTex.

Here are an empty run card and a filled one.

Share your run

Once you have built the card, it's time to convert your run to the RanxHub format and share it with the community.

First, load the run into ranx.
If you are not familiar with ranx, here you can find more information about loading your data.

Once imported, you can proceed with the conversion process as follows:

from ranx import Run 

run = Run(...)
run.save("path/to/run.lz4")

It is important to save your run in LZ4 format to lower memory footprint. ranx automatically infers the file output kind, so you just need to add ".lz4" at the end of the filename.

We encourage you to submit runs accounting for 1000 results per query.

Once finished, you are ready to share your run and its card with us by uploading them with this form. We will then upload your run on our servers and add the model card to this website.

Pretty easy, right? :)