fzliu

40 ms·

751.
Hierarchical Navigable Small Worlds(zilliz.com)
1 points by fzliu 3y ago|0 comments
752.
by fzliu 3y ago
Encoder-decoder (attention) architectures still have a tough time with long-range dependencies, so even with longer context lengths, you'll still need a retrieval solution. I agree that there's probably a better solution than pure
753.
by fzliu 3y ago
Not sure what you mean by dynamic dictionary, but the embedding tree you mention is already freely available Milvus via the Annoy index.
754.
Turn Traffic Lights Green(motorbiscuit.com)
2 points by fzliu 3y ago|1 comments
755.
by fzliu 3y ago
SCaNN is coming to Milvus very soon (we've seen some demand from the community). Stay tuned.
756.
by fzliu 3y ago
We've seen Milvus used in a variety of recommender systems running in production.
757.
by fzliu 3y ago
Agreed. There's a lot of great index types out there - HNSW is incredible, but algorithms such as ScANN (and PQ) have their place in the ecosystem. Tree-based vector indexes aren't bad either, especially if we can find a way to ma
758.
by fzliu 3y ago
Human features such as faces and fingers are easy for us humans to parse because our attention is instantly drawn to them, and we are highly sensitive to "noise" for that particular domain. AI-generated content of a three-toed ani
759.
by fzliu 3y ago
Milvus ( https://milvus.io ) and Vespa ( https://vespa.ai ) are great choices if you're looking for hardened, scalable, and production-ready vector databases. We (Milvus) also have `milvus-lite` if you'd like s
760.
by fzliu 3y ago
LangChain is meant to reduce/remove the amount of boilerplate code needed to build a lot of applications with LLMs. I see your point, but I still think LangChain is useful for a particular segment of early stage developers.
761.
Hierarchical Navigable Small Worlds(zilliz.com)
5 points by fzliu 3y ago|0 comments
762.
by fzliu 3y ago
Vector databases such as Milvus are only there to help reduce/minimize hallucinations rather than get rid of them completely. Until we have a model architecture that can perform completion from the _prompt only_ rather than pre-trainin
763.
by fzliu 3y ago
The idea of injecting domain knowledge into LLMs can certainly help, but they don't fix the problem entirely. There are still plenty of opportunities to hallucinate - for example, ChatGPT still regards the phrase "LLM" to ref
764.
Ding Liren Wins 2023 FIDE World Chess Championship(chess.com)
1 points by fzliu 3y ago|0 comments
765.
by fzliu 3y ago
Zilliz | Onsite/Hybrid (Redwood City, CA, USA) - https://zilliz.com/ Zilliz is the company behind Milvus, the world's most popular open-source vector database. We're hiring engineers, developer advocates, pro
766.
The Prisoner Who Revolutionized Language with a Teacup(wired.com)
1 points by fzliu 4y ago|0 comments
767.
by fzliu 4y ago
Even with an incredibly long context window (say, 1M tokens), attention still suffers from a problem with long-term dependencies. This is probably why OpenAI hasn't publicly released their 32k token length model just yet.
768.
Fide World Chess Championship 2023, Round 9(chess.com)
1 points by fzliu 4y ago|0 comments
769.
Flash Attention(arxiv.org)
2 points by fzliu 4y ago|0 comments
770.
Show HN: Reduce your GPT bill via semantic caching(github.com)
1 points by fzliu 4y ago|0 comments
771.
Segment Everything Everywhere All at Once(arxiv.org)
3 points by fzliu 4y ago|0 comments
772.
Reduce your GPT bill via semantic caching(github.com)
3 points by fzliu 4y ago|0 comments
773.
by fzliu 4y ago
Here's a page on the architecture of a distributed vector database (Milvus) for anybody interested: https://milvus.io/docs/architecture_overview.md
774.
Reduce your GPT bill via semantic caching(github.com)
3 points by fzliu 4y ago|0 comments
775.
by fzliu 4y ago
Certainly. Every cache (including a semantic cache like GPTCache) needs an eviction policy to prevent the size of the cache from exploding. A single line is the smallest unit a cache can hold - for a semantic cache, that means a single pair
776.
Reduce your GPT bill via semantic caching(github.com)
4 points by fzliu 4y ago|2 comments
777.
by fzliu 4y ago
OpenAI may not be training GPT-5, but Sam didn't say anything about GPT-4.5.
778.
Honda Keyfob Vulnerability (2022)(thedrive.com)
21 points by fzliu 4y ago|2 comments
779.
by fzliu 4y ago
The current implementations use Milvus and FAISS, but we have opens for other open-source vector databases as well.
780.
Show HN: GPTCache – Redis for LLMs(github.com)
7 points by fzliu 4y ago|5 comments

More ›