Sunanda Gamage

PhD candidate at Western University. Interpretable machine learning.

[Sunanda Gamage @gamage_sunanda]

@rishabh16_ @marksaroufim Visual diagrams for all ops (where applicable) would significantly cut down the time it takes to understand them. Same goes for many high level modules/ tutorials as well

[Andrej Karpathy @karpathy]

Have you ever wanted to train LLMs in pure C without 245MB of PyTorch and 107MB of cPython? No? Well now you can! With llm.c:
https://t.co/PoGTZIwASL

To start, implements GPT-2 training on CPU/fp32 in only ~1,000 lines of clean code. It compiles and runs instantly, and exactly matches the PyTorch reference implementation.

I chose GPT-2 to start because it is the grand-daddy of LLMs, the first time the LLM stack was put together in a recognizably modern form, and with model weights available.

[Gustaf Ahdritz @gahdritz]

Our preprint on OpenFold, our trainable reproduction of AlphaFold2, is finally up (https://t.co/3EoTzE3Xdb)! Since we open-sourced parameters in June, we've trained the model to high accuracy more than 50 times, on a variety of datasets. Here's what we learned (a lot) -> (1/19)

[Sunanda Gamage @gamage_sunanda]

A treasure trove for software engineers.

[🔎Julia Evans🔍 @b0rk]

it's the end of the year and some people have education budgets to spend. What tech books have really helped you?

[Sunanda Gamage @gamage_sunanda]

@visakanv Somehow, the self-hatred was never there when I was a child. I think the hyper competitive world that you step into & the expectations that pile up on you as you grow up has a lot to do with this.

[Sunanda Gamage @gamage_sunanda]

@nickcammarata And if that's the case, we can think of Pali as the language of enlightenment. Maybe a few other adjacent languages too. Attaching a language to enlightenment can be a bad idea though.

[Sunanda Gamage @gamage_sunanda]

@nickcammarata Do you feel that there are good/ direct English translations for these Pali words? I personally feel that these Pali words are loaded with context and refer to very specific states of mind (and other things) that don't translate very well to English or other languages.

[Sunanda Gamage @gamage_sunanda]

It's great to see this kind of top notch work in the natural sciences being fully open source. I hope this becomes the norm.

[Demis Hassabis @demishassabis]

Last year we presented #AlphaFold v2 which predicts 3D structures of proteins down to atomic accuracy. Today we’re proud to share the methods in @Nature w/open source code. Excited to see the research this enables. More very soon!

https://t.co/6uiV51Xly5
https://t.co/CLo7EKubBT

[Philip Vollet @philipvollet]

applied-ml curated papers, articles, and blogs on data science & machine learning in production.

@eugeneyan

https://t.co/v7Y6jVw5yI

[Sunanda Gamage @gamage_sunanda]

@Hasinirw Congratulations!

[𝙷𝚒𝚖𝚊 𝙻𝚊𝚔𝚔𝚊𝚛𝚊𝚓𝚞 @hima_lakkaraju]

If you have less than 3 hours to spare & want to learn (almost) everything about state-of-the-art explainable ML, this thread is for you! Below, I am sharing info about 4 of our recent tutorials on explainability presented at NeurIPS, AAAI, FAccT, and CHIL conferences. [1/n]

[Google DeepMind @GoogleDeepMind]

With EigenGame, our team formulated PCA as a competitive multi-agent game, testing a new approach to an old problem. @iclr_conf they shared new insights & algorithms that scale to massive data sets & offer an alternative to explore fundamental ML problems: https://t.co/NbQG9xCDnA

[Santiago @svpino]

A little over 12 years ago, the police started building a case against me.

That was stressful. They were watching. They wanted to take me off the streets.

Here is the story of how I fled Cuba and came to the United States.

↓

[Balaji @balajis]

If you’ve been fortunate enough to do well this year, consider joining me and @VitalikButerin by donating at the addresses below.

But if all you have is Twitter, help spread the word. For every RT, I’ll donate another $50 to fight COVID in India, up to $100k. #cryptovscovid

[Sunanda Gamage @gamage_sunanda]

A great read for machine learning and statistics folks.

[rob tibshirani @robtibshirani]

With postdoc Stephen Bates and Trevor Hastie, I have just completed a new paper "Cross-validation: what does it estimate and how well does it do it?" https://t.co/LrdFr6Zhbl

[Ruwandi Kariyawasam, PhD, FCCM, D(ABMM) @ruwandi_k91]

Are you an undergraduate student in Sri Lanka interested in pursuing graduate studies abroad? @goasksef is offering a 6-month mentorship program with individuals in STEM from all around the world 🌎 ‼️Deadline March 14! ‼️

[Sunanda Gamage @gamage_sunanda]

A great thread on differences between scientific fields and why there is no one scientific method.

[Kareem Carr, Ph.D. @kareem_carr]

I've worked in a lot of different sciences and what I've discovered is that each science is its own slightly bizarre alternate reality where the scientific method turned out differently.

[Sunanda Gamage @gamage_sunanda]

@KayeEllen17 London, Ontario CA. And also Colombo, Sri Lanka. I am so sorry for your loss, and I wish her soul peace.

[Sunanda Gamage @gamage_sunanda]

There are a lot of things we can learn from this pandemic about personal and public health, and I hope we carry them over to post pandemic life.

[Sabreen /صبرىن @dakhterakhter]

I’ll never forget doing compressions in a young child who was slowly dying from Influenza B, his heart being the last organ to fail. He made it (🤲🏽❤️), which speaks to his absolute tenacity.

Look at flu deaths in kids this yr. It’s remarkable. 👀

I hope we can learn from this.

[Sunanda Gamage @gamage_sunanda]

@iamtrask When you said "from a safety standpoint", what immediately came to mind was that narrow AI may be less vulnerable to attacks, more performant and robust (at the narrow task), easier to verify, more interpretable etc. than general AI.