Óscar Barquero Pérez retweeted
Si España gana su segunda Copa del Mundo el próximo domingo sortearé MIL QUINIENTOS DÓLARES (1500$) entre todos los seguidores que hagan RT.
Óscar Barquero Pérez retweeted
Si España se clasifica para la final de la Copa del Mundo sortearé MIL DÓLARES entre todos los seguidores que hagan RT.
42 grados al sol a las 17.33 y el @AytoLeganes permite celebrar la competición del pepinito. La medida ha sido retrasar una hora el inicio, en vez de a las 17.25 a las 18.25. Entiendo que si hay alguna fatalidad alguien se responsabilice
Está semana en clase tuve un episodio desasosegante. En la @urjc tenemos que repetir elecciones a rector porque contando el voto en blanco ningún candidato llegaba al 50%. En clase comenté: "os imagináis que pasa lo que en ensayo sobre la lucidez?"
Desasosiego máximo por todo mi ser. No me malinterpretéis, son chicos y chicas fantásticos, super trabajadores, muy comprometidos y colaboradores. Solo han tenido un poco de mala suerte, hasta ahora.
A partir de hoy empezaré la clase (un par de minutos)
Óscar Barquero Pérez retweeted
🚀 Transforming Decision-Making with Conformal Prediction: The Venn-Abers Revolution! 🚀
🔍 A groundbreaking paper, 'Conformal Prediction Sets Improve Human Decision Making', echoes what I’ve been advocating for a while: In the realm of business, it's not just about having well-calibrated sets, but about actionable, probability-based decisions.
Why Venn-Abers? Because it's not just about prediction; it's about precision in probability calibration for each decision class. 🎯
📚 Dive deeper into this transformative approach with my book, 'Practical Guide to Applied Conformal Prediction in Python', and discover how to apply the finest uncertainty frameworks to real-world scenarios. Get your copy here: [a.co/d/he2cuWz]
👩💻 Want to master Venn-Abers? My Medium article 'How to calibrate your classifier in an intelligent way using Venn-Abers Conformal Prediction' is your go-to guide. Read it here: [medium.com/@valeman/how-to-c…]
⚡️ 2024 is all about making any classifier smarter with Venn-Abers, and it’s easier than you think. Thanks to Ivan Petej (and his collaboration with Vovk and Valentina Fedorova [Yandex]), this method is now accessible to all. Check out the 'Venn-Abers' library: [github.com/ip200/venn-abers]
🌟 Also, don't miss Carl McBride Ellis, PhD's award-winning Kaggle notebook for a hands-on experience with Venn-Abers: [kaggle.com/code/carlmcbridee…]
#calibration #conformalprediction #datascience #ai #machinelearning
Óscar Barquero Pérez retweeted
Acoso a la bici de @madrid y @policiademadrid a tope:
ME ACABAN MULTAR 500€ x circular según las normas. La señora policía no se sabe la ordenanza, le he explicado cómo pueden circular las bicis en calles residenciales según art.169 de la ordenanza y ha pasado total! 🤯🤯🤯
¿Cuando un policía multa x un uso permitido de la vía, ignorando la normativa, se lo explicas y pasa de todo, no es prevaricación???? Tengo el audio. Es la primera multa q me ponen en la vida, ¿ahora qué se hace?
Es muy triste. No soy el primero a quien le pasa exactamente este caso en Madrid, pero me gustaría ser el último!!!! 🥲🥲🥲 – at Madrid, Spain
Óscar Barquero Pérez retweeted
This is insane! 😱
You can now train a 100-billion parameters LLM on Google Colab.
Explanation and code below ⤵️⤵️⤵️
--- TL;DR ---
Distributed training over the Internet has become operational with the release of the new version of the PETALS distributed training package. Join forces with people from around the world and receive immense computing power in return for your use! 🌐⚡️
--- Intro ---
Distributed training over the internet has turned from a dream into a reality with the release of the new version of the Petals distributed training package. Upon connection, the package grants you access to immense computing power that allows you to do anything you can imagine. It doesn't even have to be your own computing power: Attaching below is a Colab notebook that trains a 65-billion-parameter model (and also supports 100 billion. Yes, models with a hundred billion parameters are accessible to you). 💻🚀
--- It surely takes years, doesn't it? ---
No. For example, with a 65-billion-parameter inference, you'll get a speed of 5 tokens per second. ⏱️⚡️
--- How does it work? ---
The concept is very simple. In LoRA-based training (My tweet about LoRA - nitter.cf/ItakGol/status/1…), the base model is the same for everyone and does not change until the end of training, so many people need it at the same time. With Petals, people load small parts of the model, and each executes their part in the computation of running the full model.
All the headache and complex software engineering required to make this madness run smoothly and quickly happens inside the package. 🤯🔧
--- How complex? ---
From your own perspective, simply run:
AutoDistributedModel.from_pretrained
--- Important! Privacy! ---
Obviously, this infrastructure sends your data outside to the world, so it's crucial to take that into account. Your code is never sent outside. 🌐🔒
--- Links ---
1. Article: Petals: Collaborative Inference and Fine-tuning of Large Models - aclanthology.org/2023.acl-de…
2. See the speed for yourself: Distributed Chatbot - chat.petals.ml/
3. Notebook for training a 65-billion-parameter model: colab.research.google.com/dr…
4. Code: github.com/bigscience-worksh…
5. Paper: arxiv.org/abs/2209.01188
6. Join the community: discord.com/invite/D9MwApKgW…
Óscar Barquero Pérez retweeted
Happy birthday, transformer! An awesome summary @DrJimFan!
Also interesting to think about why we needed attention for RNNs (before transformers) in the first place.
Since we can't translate word-by-word, we needed a RNN encoder-decoder setup. But then, it's hard to remember.
Today 6 years ago, "Attention is All You Need" went on Arxiv! Happy birthday Transformer! 🎂
Fun facts:
- Transformer did not invent attention, but pushed it to the extreme. The first attention paper was published 3 years prior (2014) and had an unassuming title: "Neural Machine Translation by Jointly Learning to Align and Translate", from Yoshua Bengio's lab.
It is a combination of RNN + "context vectors" (i.e. attention). Many of you likely haven't heard about this paper, but it's one of the greatest milestones in NLP and has been cited 29K times (compared to Transformer's 77K).
- Neither Transformer nor the original attention paper talked about the general-purpose sequence computer. Instead, both were conceived as solutions to one narrow & specific problem: machine translation. It's remarkable that AGI (some day soon) can trace its origin to the humble Google Translate. 😅
- Transformer was published at NeurIPS 2017, one of the top AI conferences worldwide. Yet it didn't even get an Oral presentation, let alone awards. There were 3 best papers at NeurIPS that year. Combined, they have 529 citations as of today.
Óscar Barquero Pérez retweeted
Ven a conocer los grados en Ingeniería que puedes estudiar con nosotros.
Visitarás los laboratorios, conocerás a profesores y alumnos.
Te esperamos en la Jornada de Puertas Abiertas.
Inscríbete:
docs.google.com/forms/d/e/1F…
Óscar Barquero Pérez retweeted
Esto es horrible. Un informe acaba de revelar que Rusia ha enviado a miles de menores ucranianos a campos de reeducación rusos. Exijamos su regreso sanos y salvos, ¡ya! 🇺🇦 fb.avaaz.org/campaign/es/ukr…
Óscar Barquero Pérez retweeted
#1: Academic Writing Amplified
@cathymazak
• Hosted by academic writing coach Dr Cathy Mazak
• Theme Writing as a tool to advance your career
• Focus on womxn
I have been a guest, check it out 👇
podcasts.apple.com/us/podcas…
Óscar Barquero Pérez retweeted
Here is a great explanation of the Lagrangian multiplier (the intuition of which is typically not given). 🧵
The problem: maximising a function f(x,y) under a constraint g(x,y)=k (the point has to be on the red curve).
Gili-corners en el fútbol actual. Necesito ayuda @pedritonumeros @2010MisterChip. Abro pespunte 🧵
Creo que sacar siempre de una forma es completamente perjudicial, pero como nuestros propietarios saben de análisis de datos, quizá han estudiado esto y han concluido que lo mejor es siempre sacar gili-corner. me gustaría saberlo para ahorrarme el mini cabreo en cada partido