Top all-time · #18856
Songlin Yang
Personal blog · sustcsonglin.github.io/blogaiprogramming
Songlin Yang writes blank, a personal blog at sustcsonglin.github.io about ai and programming. Since 2024, 4 different Hacker News readers have linked it 6 times, most often “DeltaNet Explained (Part 1)”. It is still writing: the last post was on 2024-12-03, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link this DeltaNet-2 post to explain the chunkwise parallel algorithm, which combines parallel subsequence computation with recurrent updates between chunks for fast RNN kernels.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
A simple, whitespace theme for academics. Based on [*folio](https://github.com/bogoli/-folio) design.
Most-linked posts?
- DeltaNet Explained (Part 1)3 people · 4 mentions · first linked 2024-12-13
What Hacker News says?
Have you explored chunkwise parallel approaches? They use the O(n log n) parallel algorithm within a subsequence and update bewteen chunks recurrently like the O(n) recurrent algorithm. These are usually the fastest kernels for these kinds of RNNs. Here is the best explanation I have seen: https://sustcsonglin.github.io/blog/2024/deltanet-2/
Comment linking /blog/2024/deltanet-2/
on “Were RNNs all we needed? A GPU programming perspective”Submitted DeltaNet Explained (Part 1) as a story · 2 points
HN discussionTypically the fastest approaches for associative RNNs combine the advantages of the parallel O(n log n) algorithm with a recurrent non-parallel O(n) approach by computing results for subsequence chunks in parallel and moving to the next chunk in a recurrent manner. This blog post explains the method (chunkwise parallel algorithm) well: https://sustcsonglin.github.io/blog/2024/deltanet-2/
Comment linking /blog/2024/deltanet-2/
on “Were RNNs all we needed? A GPU programming perspective”Submitted DeltaNet Explained as a story · 2 points
HN discussionSubmitted DeltaNet Explained as a story · 1 points
HN discussionSubmitted DeltaNet Explained (Part I) as a story · 1 points
HN discussion