Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
lunaticd
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
1.
▲
by
lunaticd
2y ago
3. The project started under a Harvard affiliated Github org during the course of PhDs. These same people later joined Google where it continued to be developed and over time adopted more and more in place of TensorFlow.
2.
▲
Optimizing RAG to Search PubMed Articles in Milliseconds Instead of Minutes
(domluna.com)
2 points
by
lunaticd
2y ago
|
0 comments
3.
▲
Approximate Binary Vector Search for RAG in Julia
(domluna.com)
1 points
by
lunaticd
2y ago
|
0 comments
4.
▲
by
lunaticd
2y ago
i did it does a ton of allocations, which is why i made a simple maxheap implementation and then sort the final result
5.
▲
by
lunaticd
2y ago
please make it even faster!
6.
▲
by
lunaticd
2y ago
it doesn't seem to have better support for things like xor and count_ones. I believe the main use case is comparisons.
7.
▲
by
lunaticd
2y ago
julia> @code_typed hamming_distance(Int8(33), Int8(125)) CodeInfo( 1 ─ %1 = Base.xor_int(x1, x2)::Int8 │ %2 = Base.ctpop_int(%1)::Int8 │ %3 = Base.sext_int(Int64, %2)::Int64 │ nothing::Nothing └── return %3 ) => Int64
8.
▲
by
lunaticd
2y ago
exact in this case means that all the vectors are compared against the query vector. Where as other search methods such as HNSW are approximate searches.
9.
▲
by
lunaticd
2y ago
author here. I thought there might be a machine instruction for this but wasn't sure, I also didn't know Julia had a count_ones that counted the 1s. Thanks! With this the timings are even faster. I'll update the post.
10.
▲
Exact binary vector search for RAG in 100 lines of Julia
(domluna.com)
180 points
by
lunaticd
2y ago
|
23 comments
11.
▲
Generative AI vs. AWS Textract – A Case for GPT Wrappers
(domluna.com)
1 points
by
lunaticd
3y ago
|
1 comments
12.
▲
LLMs can dream of electric sheep, but can they add?
(domluna.com)
4 points
by
lunaticd
3y ago
|
0 comments
13.
▲
Take1 – An Exploration into LLM Writing
(domluna.com)
1 points
by
lunaticd
3y ago
|
0 comments
14.
▲
Counting NanoGPT FLOPs in PyTorch
(domluna.com)
1 points
by
lunaticd
3y ago
|
0 comments
15.
▲
Show HN: DataGPTd – Making data analysis as easy as asking a question
(datagptd.com)
2 points
by
lunaticd
3y ago
|
0 comments
16.
▲
Grokking
(domluna.com)
122 points
by
lunaticd
3y ago
|
29 comments
17.
▲
TableChopper: Edit CSVs with Natural Language
(tablechopper.pages.dev)
8 points
by
lunaticd
4y ago
|
2 comments