From 1e8fa44a73b2daa0bd6dbf9fe28801a6b96bb8ed Mon Sep 17 00:00:00 2001 From: Sabrina Aquino <77522207+sabrinaaquino@users.noreply.github.com> Date: Thu, 26 Sep 2024 12:50:18 -0300 Subject: [PATCH] adding slower indexing to PQ on when to use section --- qdrant-landing/content/articles/what-is-quantization.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/qdrant-landing/content/articles/what-is-quantization.md b/qdrant-landing/content/articles/what-is-quantization.md index f46dfbbf8..bc570c297 100644 --- a/qdrant-landing/content/articles/what-is-quantization.md +++ b/qdrant-landing/content/articles/what-is-quantization.md @@ -515,7 +515,7 @@ Here are some final thoughts to help you choose the right quantization method fo |--------------------------|-------------------------------------------------------------|--------------------------------------------------------------------------------------------| | **Binary Quantization** | • **Fastest method and most memory-efficient**
• Up to **40x** faster search and **32x** reduced memory footprint | • Use with tested models like OpenAI's `text-embedding-ada-002` and Cohere's `embed-english-v2.0`
• When speed and memory efficiency are critical | | **Scalar Quantization** | • **Minimal loss of accuracy**
• Up to **4x** reduced memory footprint | • Safe default choice for most applications.
• Offers a good balance between accuracy, speed, and compression. | -| **Product Quantization** | • **Highest compression ratio**
• Up to **64x** reduced memory footprint | • When minimizing memory usage is the top priority
• Acceptable if some loss of accuracy is tolerable | +| **Product Quantization** | • **Highest compression ratio**
• Up to **64x** reduced memory footprint | • When minimizing memory usage is the top priority
• Acceptable if some loss of accuracy and slower indexing is tolerable | ### Learn More @@ -528,4 +528,4 @@ Learn more about optimizing real-time precision with oversampling in Binary Quan -Stay up-to-date on the latest in [vector search](/advanced-search/) and quantization, share your projects, ask questions, [join our vector search community](https://discord.com/invite/qdrant)! \ No newline at end of file +Stay up-to-date on the latest in [vector search](/advanced-search/) and quantization, share your projects, ask questions, [join our vector search community](https://discord.com/invite/qdrant)!