3 ms·
Love it! Need better ANNs approaches in databases, using external servers such as Milvus, faiss is a pita related: https://github.com/netrasys/pgANN https://gi
by phenkdo 5y ago
Love it! Need better ANNs approaches in databases, using external servers such as Milvus, faiss is a pita
related: https://github.com/netrasys/pgANN https://github.com/netrasys/pgANN
- gk1 5y agoWhy are they a pain in the...?
- phenkdo 5y agothe usual yet-another-moving-part (YAMP) complexity.
- EmilStenstrom 5y agoThey typically load all data in memory, so you still need persistence to handle crashes (two setups). And since data is typically huge you need servers with lots of expensive RAM.
- patelajay285 5y agoCheck out Magnitude, we built it to solve that problem: https://github.com/plasticityai/magnitude https://github.com/plasticityai/magnitude It's still loaded from a file, but heavily uses memory-mapping and caching to be speedy and not overload your RAM immediately. And in production scenarios, multiple worker processes can share that memory due to the memory mapping. Granted it's read-only, so might not be exactly what you are looking for. Disclaimer: I'm the author.
- rkt08 5y agoHow about a vector oriented 'database' instead? Pinecone(https://www.pinecone.io/ https://www.pinecone.io/) does both exact and approx search and it's fully managed so you don't have to worry about reliability,availability etc. PS: I work there