ds

[Aritra 🤗 @ariG23498]

When you run a @PyTorch model on a GPU, the acutal work is executed through kernels. These are low-level, hardware-specific functions designed for GPUs (or other accelerators).

If you profile a model, you'll see a sequence of kernel launches. Between these launches, the GPU can sit idle, waiting for the next operation. A key optimization goal is therefore to minimize gaps between kernel execution and keep the GPU fully utilized.

One common approach is `torch.compile`, which fuses multiple operations into fewer kernels, reducing overhead and improving utilization.

Another approach is to write custom kernels tailored to specific workfloads (e.g., optimized attention or fused ops). However, this comes with significant challenges:

> requires deep expertise in kernels writing
> installation hell
> integration with the model is non-trivial

To address this,@huggingface introduces the `kernels` library.

With this one can:
> build custom kernels (with the help of a template)
> upload them to the Hub (like models or datasets)
> integrate them to models with ease

Let's take a look at how the transformers team use the kernels library to integrate it into the already existing models. (more in the thread)

[François Chollet @fchollet]

Sufficiently advanced agentic coding is essentially machine learning: the engineer sets up the optimization goal as well as some constraints on the search space (the spec and its tests), then an optimization process (coding agents) iterates until the goal is reached.

The result is a blackbox model (the generated codebase): an artifact that performs the task, that you deploy without ever inspecting its internal logic, just as we ignore individual weights in a neural network.

This implies that all classic issues encountered in ML will soon become problems for agentic coding: overfitting to the spec, Clever Hans shortcuts that don't generalize outside the tests, data leakage, concept drift, etc.

I would also ask: what will be the Keras of agentic coding? What will be the optimal set of high-level abstractions that allow humans to steer codebase 'training' with minimal cognitive overhead?

[Neural Machine @neural_machine]

Волк не показывает, что он голоден, потому что он уже поел.

[Neural Machine @neural_machine]

Когда волк в тепле, он слаб.

[Noam Brown @polynoamial]

Social media tends to frame AI debate into two caricatures:
(A) Skeptics who think LLMs are doomed and AI is a bunch of hype.
(B) Fanatics who think we have all the ingredients and superintelligence is imminent.

But if you read what leading researchers actually say (beyond the headlines), there’s a surprising amount of convergence:

1) The current paradigm is likely sufficient for massive economic and societal impact, even without further research breakthroughs.

2) More research breakthroughs are probably needed to achieve AGI/ASI. (Continual learning and sample efficiency are two examples that researchers commonly point to.)

3) We probably figure them out and get there within 20 years. @demishassabis said maybe in 5-10 years. @fchollet recently said about 5 years. @sama said ASI is possible in a few thousand days. @ylecun said about 10 years. @ilyasut said 5-20 years. @DarioAmodei is the most bullish, saying it's possible in 2 years though he also said it might take longer.

None of them are saying ASI is a fantasy, or that it's probably 100+ years away.

A lot of the disagreement is in what those breakthroughs will be and how quickly they will come. But all things considered, people in the field agree on a lot more than they disagree on.

[Mike @MikeLongTerm]

BREAKING Investors discovered $GOOGL TPU after 10 years.

And investors also discovered @Google uses TPU to train its models since 2016 without $NVDA for almost 10 years is shocking.

[Fermat's Library @fermatslibrary]

The reason why Olympus Mons is the largest volcano in the Solar System is because Mars doesn't have plate tectonics, allowing magma to rise and pile up. Low gravity and low erosion also help.

[Historyland @HistorylandHQ]

Roman mosaic showing a tigress with young tigers. Object dated to the 4th century CE. It is now at The Cleveland Museum of Art in the USA. [1200x1312]

[Sam Altman @sama]

no thank you but we will buy twitter for $9.74 billion if you want

[Martin Shkreli @MartinShkreli]

i heard the US sovereign wealth fund needs a biotech PM

[ds @desanzy]

Rex nunquam moritur

[Neural Machine @neural_machine]

Если хочешь заниматься любовью, люби себя

[. @StockCats_2009]

these new dollar bills are weird, it's going to take some time to get used to them

[Forbes Russia @ForbesRussia]

В чистом и аккуратном Минске самые желанные жители — сотрудники высокотехнологичных компаний. Их здесь уже 60 000, и цифра эта стремительно растет. Forbes разобрался в белорусском цифровом чуде
https://t.co/rEhTPJUnvz

[Френдли Фраер @Venya_sick]

Сделал вам меме

[Alex Kontorovich @AlexKontorovich]

With Pi Day just around the corner, let’s remember what Pi is all about.

After washing your hands thoroughly, cut the crust off a pizza pie and lay it across four others. You’ll see that the crust spans a little more than 3 pies. That’s Pi ≈ 3.14.

But that’s not all! (Cont’d)

[Lisa Abramowicz @lisaabramowicz1]

The more panic in markets, the more traders amp up bets for Fed rate cuts, with almost 3 cuts fully priced in by September. The idea of a Fed put, however, isn't helping risk appetite much, which raises a question about how much the Fed can stimulate from here.

[Reuters @Reuters]

Russian tennis star Maria Sharapova, five-time Grand Slam winner, retires https://t.co/aovi6AbfaW

[Сноб @snob_project]

Первые впечатления о городе обычно остаются с нами навсегда. Татьяна Абрамовская долгое время прожила в Париже, она делится воспоминаниями о первых днях в столице Франции
https://t.co/rQxvrrSgvu