Show HN: A fast HNSW implementation in Rust
1–10 of 30 posts
Re: Show HN: A fast HNSW implementation in Rust
#2How does this compare to hsnwlib - is it faster? https://github.com/nmslib/hnswlib
Re: Show HN: A fast HNSW implementation in Rust
#3Hello, I have a few questions:
- how much time to insert 15 millions of vectors of 768 f32?
- how much RAM needed for this operation?
- if inserting another vector, how incremental is the insertion? Is it faster than reindexing the 15M + 1 vectors from scratch?
- does the structure need to stay in RAM or can it be efficiently queried from a serialized représentation?
- how fast is the search in the 15M vectors on average?
Re: Show HN: A fast HNSW implementation in Rust
#4How does this compare to hsnwlib - is it faster? https://github.com/nmslib/hnswlib
Also compare with qdrant's Rust implementation; they tout their performance. https://github.com/qdrant/qdrant/tree/master/lib/segment/src...
Re: Show HN: A fast HNSW implementation in Rust
#5As a Rust dev myself, can we stop the trend of adding "in Rust" to titles?
The loud part should be the thing that was built, not the thing it was built with.
That said, this is super cool. I have a project that I can definitely benefit from this. :)
Re: Show HN: A fast HNSW implementation in Rust
#6As a Rust dev myself, can we stop the trend of adding "in Rust" to titles? The loud part should be the thing that was built, not the thing it was built with. That said, this is super cool. I have a project that I can definitely benefit from this. :)
As a pythonista and wannabe Rustaceans, I personally love seeing high performance tooling built in Rust!
Re: Show HN: A fast HNSW implementation in Rust
#7Great job shipping. The README / repo however could definitely benefit with a benchmarking report in them.
Re: Show HN: A fast HNSW implementation in Rust
#8Hello, I have a few questions: - how much time to insert 15 millions of vectors of 768 f32? - how much RAM needed for this operation? - if inserting another vector, how incremental is the insertion? Is it faster than reindexing the 15M + 1 vectors from scratch? - does the structure need to stay in RAM or can it be efficiently queried from a serialized représentation? - how fast is the search in the 15M vectors on ave…
I can answer #3. HNSW will allow for incremental index rebuilding. So each additional insert is a sublinear, but greater than constant time, operation.
Re: Show HN: A fast HNSW implementation in Rust
#9As a Rust dev myself, can we stop the trend of adding "in Rust" to titles? The loud part should be the thing that was built, not the thing it was built with. That said, this is super cool. I have a project that I can definitely benefit from this. :)
> As a Rust dev myself, can we stop the trend of adding "in Rust" to titles?
tbf it's the headline of the Readme.
More generally I largely agree for software. I don't really agree for library code like this though... I actually care about what language a library is implemented in when I'm coding up a project that might use it.
Re: Show HN: A fast HNSW implementation in Rust
#10Happy 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-rust/blob/b8ef946bd76112250...) and so is the base layer beam search.