leod

6 ms·

1.
by leod 3y ago
Happy to see people working on vector search in Rust. Keep it up! As far as HNSW implementations go, this one appears to be almost entirely unfinished. Node insertion logic is missing ( https://github.com/swapneel/hnsw-r
2.
by leod 3y ago
This is a fair point and also called out in the discussion section. To some degree, this could be mitigated by hotloading shader code (and compiling shaders in debug mode). However, this remains as a fundamental downside of the approach. Pe
3.
Posh: Type-Safe Graphics Programming in Rust(leod.github.io)
20 points by leod 3y ago|10 comments
4.
by leod 4y ago
Awesome work. Would you be willing to share details about the fine-tuning procedure, such as the initialization, learning rate schedule, batch size, etc.? I'd love to learn more. Background: I've been playing around with generatin
5.
by leod 7y ago
Interesting. They train an image classifier to detect images that were generated by a GAN-trained CNN. I wonder if it could be possible to include this classifier in the training loss, such that the generated images fly under its radar as m
6.
by leod 7y ago
Ha! In the model's defense, its training data [1] ends in 2017 -- not sure if hatred for Electron was as prevalent back then. [1] https://archive.org/details/14566367HackerNewsCommentsAndSto...
7.
by leod 7y ago
Thanks! I actually planned to make results shareable at the start, but, knowing the internet, I did not like the idea of being held responsible for whatever content (say offensive or even illegal things) people would put into the titles.
8.
by leod 7y ago
I've trained a Transformer encoder-decoder model (this was slightly before GPT2 came out) to generate HN comments from titles. There is a demo running at https://hncynic.leod.org
9.
Show HN: A bot that tweets generated comments for the HN front page(twitter.com)
4 points by leod 7y ago|0 comments
10.
by leod 7y ago
Thank you so much for your comprehensive answer, this helps a lot. If I understand nshepperd's code correctly, it uses a constant and small learning rate. Do you know if this works better than the learning rate schedule that is usually
11.
by leod 7y ago
I haven't heard of gradient checkpointing yet, thank you for the link! Do you know how it compares to gradient accumulation? The latter basically reduces the batch size, but takes the sum of multiple gradients before actually performin
12.
by leod 7y ago
If humans are not limited by the halting problem, it would be great if you could tell me if the following function f halts for all integers n: https://gist.github.com/leod/9b89af30cff21cb925d4522a68c990d...
13.
by leod 7y ago
Thank you! The model weighs in at 1.2GB with 100M parameters, which is similar to the smallest GPT-2 model. I wouldn't be suprised if GPT-2 small (+ finetuning on HN data) performed better than what I have trained. Other than hyperpara
14.
Show HN: hncynic – Generate Hacker News Comments from Titles(hncynic.leod.org)
114 points by leod 7y ago|28 comments
15.
by leod 8y ago
This is true for some scenarios, like invalidating iterators through deletion (detected at compile-time by the borrow checker), but other scenarios still require runtime checks, right? Consider e.g. array out of bound accesses -- are you aw
16.
by leod 9y ago
This analysis seems to be based on some notion of "Cognitive Complexity", but I can't find its definition in the article. Am I missing something?
17.
by leod 9y ago
True, but that does not show that the approximation can always be learned from examples, does it?
18.
by leod 11y ago
That's certainly an interesting idea, but wouldn't some people just "reply all" for more visibility?
19.
by leod 13y ago
It certainly struck me as an odd analogy. There are plenty of theories that are hard to disprove, why choose one as politically loaded as "Marxism"?
20.
by leod 13y ago
I quite enjoyed The User Illusion (the author has a great way of looking at the world). I think its chapters on consciousness are mostly about what consciousness gets to see. Building up from the idea that there is a very limited amount o
21.
by leod 13y ago
I'm a Vim user as well, and when I tried Colemak for a few months some years back, Vim was what made me go back to QWERTY. It was just too awkward as I would like to be able to work with both layouts. Creating new bindings didn't
22.
by leod 14y ago
Maybe they believe that there is a higher long-term profit in having well-educated workers.
23.
by leod 14y ago
This is, of course, assuming, that the government is creating this law for the reason of solving crime and not for some special interests / lobbies.
24.
by leod 14y ago
Saying that the modern SPD is leftwing is like saying that the Democrats in the USA are leftwing.