Could the next #nlp tutorial please please please please NOT be sentiment analysis? Surely there are other binary classifiers to teach? or this might be far fetched… multi class classification? 🤯😱

🤩

Don't know when I started doing this but it is great to understand smaller pieces of a larger program by creating toy projects (experiments). It removes confounding clutter and context so you can focus on the exact problem in a much much sharper way.

Really enjoying the detailed code along walkthroughs of Hugging Face with "rupert ai" on YouTube youtube.com/watch?v=vNKIg8… #nlp #ml

After 2 years, Practical Deep Learning for Coders v5 is finally ready! 🎊 This is a from-scratch rewrite of our most popular course. It has a focus on interactive explorations, & covers PyTorch, Hugging Face, DeBERTa, ConvNeXt, Gradio & other goodies 🧵 course.fast.ai

Gonna start a conference called #NormIPS that’s just presentations of middlebrow ML topics. “how to structure Python packages 2022”, “how many k-folds is too many”, “how to make the browser pop-up come up when the notebook is done running”, “putting features in Postgres”, etc.

As a data professional, I am embarrassed with how little I knew about how pervasive ApacheArrow is at making my life better until now. 👏 youtu.be/YhF8YR0OEFk

At first I thought this was a new way of describing types of trends in time series data. Eg “I’m expecting interest rates to Cinderella but the economy is Man in a Hole”

So here's a short #Python story. My 14 yo son made a game for the Pew Pew console, a tiny Dance Dance Revolution clone. He learned to program the console at @deshipu's workshop at EuroPython this year. The final DDR clone was just 100 LOC and worked fine but there was a problem.

Very much enjoying reading the early release content of “Scaling #MachineLearning on Apache Spark” by Adi Polak Truly amazing to get bleeding edge books with incremental chapter releases to provide the distilled narrative around complex topics. learning.oreilly.com/library/view/s…

Probably all the async things in Python that claim high performance (throughput), including FastAPI, are just powered by this. 🌪 ✨ Uvloop ✨ is the secret ingredient underneath all this (e.g. powering Uvicorn). This is what makes Python a good alternative to NodeJS or Go. 🤓

babe are you ok? you’ve barely touched the dashboard you claimed was mission critical

new little song: "The Re-Org Rag (I'm My Own VP)"

Amazing and underrated interview with Nikita | Scaling Postgres about the future of datastores. It is a matter of time until Snowflake, Redshift and Databricks are all playing catch-up. As a consultant in Data & AI I am keeping at eye on Neon Postgres.

🤫

Python 3.11 is finally released. In the CPython release team, we have put a lot of effort into making 3.11 the best version of Python possible. Better tracebacks, faster Python, exception groups and except*, typing improvements and much more. Get it here: python.org/downloads/rele…

This is how I explain overfitting now.

The fact the we as a society have figured out how to communicate at the speed of light (fibre optics) and the speed of electricity over copper wire (1/3 speed of light) so a whole planet of distance doesn’t matter is phenomenal Now the limiting factor is emotional intelligence.

Has anyone else ever thought about how nodes in a cluster never need couples therapy or HR involvement to mediate because they have an agreed protocol and communication style? Just me?

Not all heroes wear capes. I’ve always said spreadsheets are gateway software development.