Top all-time · #6168
Horace He
Personal blog · horace.io · on HN as chilleeaiprogramming
Horace He's personal blog at horace.io covers ai and programming. Since 2021, 13 different Hacker News readers have linked it 22 times, most often “Making Deep Learning Go Brrrr from First Principles”. No feed was found, so read it on the site itself.
Too few sampled comments (6) to say how often HN recommends it.?
Why HN links it
HN readers link Horace He's posts on PyTorch versus TensorFlow adoption, including a graph of framework usage in papers, and the 'brrr' first-principles article on deep learning performance and overhead.
Summarised by Claude from a sample of 6 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- Making Deep Learning Go Brrrr from First Principles8 people · 12 mentions · first linked 2022-03-15
- PyTorch vs. TensorFlow in Academic Papers2 people · 4 mentions · first linked 2021-11-27
What Hacker News says?
Exactly. PyTorch is not only easy, but is a joy to work with. Among researchers, TensorFlow is rapidly losing ground to PyTorch, and, I think, will keep losing ground until it becomes a niche and only used by Googlers and some others. https://horace.io/pytorch-vs-tensorflow/
Comment linking /pytorch-vs-tensorflow/
on “PyTorch: Where we are headed and why it looks a lot like Julia (but not exactly)”Submitted PyTorch vs. TensorFlow in Academic Papers as a story · 103 points
HN discussionAnything that’s not an actual computational kernel (python interpreter, Pytorch dispatcher, etc.) See the flame graph here in the overhead section: https://horace.io/brrr_intro.html
Comment linking /brrr_intro.html
on “Doing small network scientific machine learning in Julia faster than PyTorch”Submitted Making Deep Learning Go Brrrr from First Principles as a story · 55 points
HN discussionInteresting decision to go with tensorflow on a new system. If you have no choice because of existing code it's understandable, though recent research has clearly shifted to pytorch [1]. Though if you're required to use tensorflow, will you even be able to switch to the bleeding edge version? N=1 but all projects I worked on where customers required tensorflow had to be on version <2 http://horace.io/pytorch-vs-tens…
Comment linking /pytorch-vs-tensorflow/
on “Are the New M1 Macbooks Any Good for Deep Learning?”Submitted How many paths of length K are there between A and B? (2021) as a story · 34 points
HN discussionTo support your point with data, here is a graph of usage of TF vs PyTorch in papers over time: https://horace.io/pytorch-vs-tensorflow/
Comment linking /pytorch-vs-tensorflow/
on “Ask HN: What ML platform are you using?”Submitted Making Deep Learning Go Brrrr from First Principles as a story · 3 points
HN discussionDirect link to article: https://horace.io/brrr_intro.html
Comment linking /brrr_intro.html
on “In the time Python can perform 1 FLOP, an A100 can chew through 9.75M FLOPS”Submitted Making Deep Learning Go Brrrr from First Principles as a story · 3 points
HN discussionThanks! I've also written a couple other things along a similar vein you might like at https://horace.io/writing.html (particularly https://horace.io/brrr_intro.html ) and also some of the things I've tweeted: https://twitter.com/cHHillee/highlights
Comment linking /brrr_intro.html
on “Accelerating Generative AI with PyTorch II: GPT, Fast”Submitted PyTorch vs. TensorFlow as a story · 3 points
HN discussionI think gp's referring to this meme [1] embedded in the article. I believe a charitable interpretation would be that it's satirising the various entities who have voiced opinions against large scale deep learning (from skeptics and theorists, to social activist types that want to slow or stop DL's march). No need to take it too seriously. [1] https://horace.io/img/perf_intro/gpus_go_brrr.webp
Comment linking /img/perf_intro/gpus_go_brrr.webp
on “Making deep learning go brrrr from first principles”Submitted Making Deep Learning Go Brrrr from First Principles as a story · 2 points
HN discussionLook, Carmack is a genius in his own corner, but he is taking that quote vastly out of context. The actual linked article[1] is quite fascinating, and does point to overhead costs as being a potential bottleneck, but that specific quote is more to do with GPUs being faster than CPUs rather than anything about Python in particular. More specifically, overhead (Python + pyTorch in this case) is often a bottleneck when…
Comment linking /brrr_intro.html
on “Why I'm still using Python”Submitted Making deep learning go brrrr from first principles as a story · 2 points
HN discussionSubmitted Making Deep Learning Go Brrrr from First Principles as a story · 2 points
HN discussionSubmitted How Many Paths of Length K Are There Between a and B? as a story · 2 points
HN discussionSubmitted How Many Paths of Length K?: and I Show You How Deep the Rabbit Hole Goes as a story · 2 points
HN discussionSubmitted Making Deep Learning Go Brrrr from First Principles as a story · 1 points
HN discussionSubmitted Making Deep Learning Go Brrrr from First Principles as a story · 1 points
HN discussion
