Hackathon #Somos600M: How to participate

• April 10th 23:59Anywhere On Earth: Deadline tosubmit projectsto the Hackathon #Somos600M and theLatinX in NLP @NAACLWorkshop.
• April 11th: Live project presentations, 5 mins per team.
• April 18th: Announcement of the winning projects.
• Soon: Live presentation of the winning projects, 30 mins per team.

All participants 👏

• Access to PRO endpoints on Hugging Face for creating synthetic corpora.
• Access to GPUs with up to 25GB of RAM on Hugging Face for model training and demo.
• Access to “persistent storage” on Hugging Face for creating Argilla annotation spaces.
• Support to present your project at the LatinX in NLP @NAACL 2024 workshop, one of the most important international NLP conferences. Learn how inthis post. Also, the LatinX in AI team is available for questions!

Everyone that presents a project 🚀

• Certificate of participation or winning team of the “Hackathon #Somos600M 2024” (verified on our website).

• 20% discount for theWomenTech Global Conference 2024.
• Possibility of obtaining a completely free pass to attend the WomenTech Global Conference 2024 (let us know your interest in the project submission form).
• Possibility of receiving a nomination to joinNova(let us know your interest in the project submission form).
• Possibility to continue developing your project with our support, contact us!

3rd place team (prizes per person) 🥉

• Certificate, recognition on the website and social media, and honorary role in the Discord server.
• 20k credits from the MonsterAPI byQ Blocksfor LLM training.

2nd place team (prizes per person) 🥈

• Certificate, recognition on the website and social media, and honorary role in the Discord server.
• 30k credits from the MonsterAPI byQ Blocksfor LLM training.

1st place team (prizes per person) 🥇

• Certificate, recognition on the website and social media, and honorary role in the Discord server.
• 50k credits from the MonsterAPI byQ Blocksfor LLM training.

A complete project consists of instruction corpora + model + demo. Likewise, given the hackathon’s focus on data, we also accept projects that have focused on corpus creation (maximum score: 7 points).

Corpus (4 points):

• Focus on linguistic varieties
• Correct corpus structure
• Corpus creation technique
• Clarity and reproducibility of scripts
• Completeness of the Dataset Card
• Corpus quality

Model (3 points):

• Training method used
• Clarity and reproducibility of scripts
• Completeness of the Model Card
• Model evaluation

Demo (1 point):

• Clarity and UX of the demo

Project and presentation (2 points):

• Motivation, originality, and social impact
• Clarity and quality of the presentation

Extra point:

• Each jury member can assign an extra point to a project that has particularly caught their attention.

Absolutely! While the focus of the hackathon is on the Spanish language and its varieties, we welcome participants from all over the world. Diversity enriches the projects and the community!

By joining this hackathon, you will have the opportunity to:

• ✅ Understand how large language models (LLMs) work and discover the challenges of each stage of their development: corpus creation, training, and evaluation
• ✅ Participate in the creation of a quality and diverse corpus that includes the different varieties of Spanish and co-official languages (top as an experience and top for the CV)
• ✅ Resolve all your doubts about NLP during “Ask Me Anything” mentoring sessions
• ✅ Receive support to present your work in a paper
• ✅ Win prizes to continue growing as a professional and get a certificate
• ✅ Join the largest community of Spanish speakers who study, work, and research in NLP

From the SomosNLP team, we want to encourage you to participate regardless of your current knowledge. In previous editions, we have had groups from research institutes and groups of undergraduate students, all projects add up!

We are at your disposal to help you in every step of the development of your project! Just post your question in the #pide-ayuda channel or ping us in your project’s thread.

We will provide an example of how to create a dataset, train a model, and create a demo. It’s up to you and your team to choose how much to research and work to improve the base version. The difficulty also depends on the use case, the origin of the data, the time you dedicate to its curation, the training technique, the iterations you make, and how elaborate you want your demo to be. You are free to choose everything!

Ideally all teams will be registered during the first week of the hackathon, until March 8th.

EDIT: We accept new teams until April 7th.

Finding a team is easy! Check the README of the #encuentra-equipo channel on our Discord server!

You have two options:

• 👀 Filter for posts from other participants who are “looking for people” and respond to them, OR
• 📢 Create a new thread specifying the topic you would like to work on

We encourage diversity in teams, including a mix of skills, experiences, and backgrounds. This diversity often leads to more innovative and comprehensive projects.

Yes, we accept teams between 1 and 5 people.

• Use your project channel on Discord to communicate and organize
• Since it’s an international hackathon, we recommend asynchronous communication or dividing the work and holding meetings with fewer people
• Schedule meetings or talk spontaneously using the new voice channels in the “SALAS DE REUNIÓN” (meeting rooms) category on Discord
• Pin important messages in the project channel, e.g., task allocation, day of the next meeting, … To pin a message, click on the three dots and select “Pin message”
• For better clarity, you can also create a shared document with team members to write down the project’s objective, allocate tasks, and more (and pin the link in the chat)