Hiiiiii Weaviate Friends!!!
There has been a lot going on in the last few weeks. We have a few exciting blogs, podcasts and stories to share with you.
Our very own Laura Ham has been researching: how we could combine the power of cross-encoders and bi-encoders to get more accurate results without sacrificing performance.
✍️ Read Using Cross-Encoders as reranker in multistage vector search - to learn about:
What cross-encoders and bi-encoders are
How to combine them
GOMEMLIMIT – a game changer for high-memory applications
The Go team have released GOMEMLIMT – an exciting feature that we particularly feel very excited about. It allows us to better control the garbage collection process, and avoid the rare out-of-memory application crashes.
✍️ Read GOMEMLIMIT is a game changer for high-memory applications - by Etienne Dilocker – to learn about:
The memory allocation process in Go
The Garbage Collector
Scenarios that could result in OOM
How GOMEMLIMIT helps
The results of our experiments
If you find this interesting, join us in the conversation on Reddit.
Integration story - BetterFeedback
We've been collaborating with our friends at BetterFeedback.ai, who in their quest to give their customers a better analytics experience, decided to join the Semantic Search movement. But don't take my word for it:
✍️ Read The Full Power of Semantic Search and Vector Database - by Rafał Muszyński - to learn about their motivations and why Weaviate was their pick.
Weaviate helps 🚀NASA's GES-DISC team
We are very excited to see Weaviate in a presentation by NASA's GES-DISC team.
👀 See GES-DISC Graph-Enabled Vector Search - where they show how they are using language embeddings and graph-enabled vector search (Weaviate) to improve their search capabilities.
Last week, Connor met with Yaoshiang Ho – a Co-Founder of Masterful AI.
🎙Watch Weaviate Podcast #22 – to learn more about:
cutting edge Computer Vision algorithms such as Noisy Student, SimCLR, and Barlow Twins
broader topic of Semi-Supervised Learning
We had an update to the Multi2Vec CLIP inference module, now on version 1.1.4 (notes)
Fix unserializable values in /meta endpoint
😍 Often, I use the 🚀 emoji to visualize how things are going with @weaviate_io and vector search. But seeing this slide (link below) in the deck "GES-DISC Graph-Enabled Vector Search" by @NASA's GES-DISC team, I feel this emoji has never been more appropriate 🙏
— Bob van Luijt (@bobvanluijt)
Aug 10, 2022