Merge pull request #628 from qdrant/mjang-convert-bq-openai-to-article

Move Binary Quantization blog post to articles
This commit is contained in:
Sabrina Aquino
2024-02-22 21:41:45 -03:00
committed by GitHub
13 changed files with 70 additions and 8 deletions
@@ -4,13 +4,13 @@ draft: false
slug: binary-quantization-openai slug: binary-quantization-openai
short_description: Use Qdrant's Binary Quantization to enhance OpenAI embeddings short_description: Use Qdrant's Binary Quantization to enhance OpenAI embeddings
description: Use Qdrant's Binary Quantization to enhance the performance and efficiency of OpenAI embeddings description: Use Qdrant's Binary Quantization to enhance the performance and efficiency of OpenAI embeddings
preview_image: /blog/openai/Article-Image.png # Change this preview_dir: /articles_data/openai/preview
preview_image: /articles-data/openai/Article-Image.png # Change this
small_preview_image: /articles_data/openai/icon.svg
social_preview_image: /articles-data/openai/preview/social-preview.png
title_preview_image: /articles-data/openai/preview/preview.webp # Optional image used for blog post title
social_preview_image: /blog/openai/Article-Image.png # Optional image used for link previews date: 2024-02-21T13:12:08-08:00
title_preview_image: /blog/openai/Article-Image.png # Optional image used for blog post title
small_preview_image: /blog/openai/Article-Image.png # Optional image used for small preview in the list of blog posts
date: 2024-02-20T13:12:08-08:00
author: Nirant Kasliwal author: Nirant Kasliwal
author_link: https://www.linkedin.com/in/nirant/ author_link: https://www.linkedin.com/in/nirant/
@@ -19,7 +19,9 @@ tags:
- OpenAI - OpenAI
- binary quantization - binary quantization
- embeddings - embeddings
weight: 0 # Change this weight to change order of posts weight: -130
aliases: [ /blog/binary-quantization-openai/ ]
--- ---
OpenAI Ada-003 embeddings are a powerful tool for natural language processing (NLP). However, the size of the embeddings are a challenge, especially with real-time search and retrieval. In this article, we explore how you can use Qdrant's Binary Quantization to enhance the performance and efficiency of OpenAI embeddings. OpenAI Ada-003 embeddings are a powerful tool for natural language processing (NLP). However, the size of the embeddings are a challenge, especially with real-time search and retrieval. In this article, we explore how you can use Qdrant's Binary Quantization to enhance the performance and efficiency of OpenAI embeddings.
@@ -199,4 +201,6 @@ We recommend the following best practices for leveraging Binary Quantization to
4. Rescoring: Enable rescoring to improve the accuracy of search results. 4. Rescoring: Enable rescoring to improve the accuracy of search results.
5. RAM: Store the full vectors and payload on disk. Limit what you load from memory to the binary quantization index. This helps reduce the memory footprint and improve the overall efficiency of the system. The incremental latency from the disk read is negligible compared to the latency savings from the binary scoring in Qdrant, which uses SIMD instructions where possible. 5. RAM: Store the full vectors and payload on disk. Limit what you load from memory to the binary quantization index. This helps reduce the memory footprint and improve the overall efficiency of the system. The incremental latency from the disk read is negligible compared to the latency savings from the binary scoring in Qdrant, which uses SIMD instructions where possible.
Want to discuss these findings and learn more about Binary Quantization? [Join our Discord community.](https://discord.gg/qdrant) Read the documentation: [Binary Quantization.](https://qdrant.tech/documentation/guides/quantization/?selector=aHRtbCA%2BIGJvZHkgPiBkaXY6bnRoLW9mLXR5cGUoMSkgPiBzZWN0aW9uID4gZGl2ID4gZGl2ID4gZGl2Om50aC1vZi10eXBlKDIpID4gYXJ0aWNsZSA%2BIGgyOm50aC1vZi10eXBlKDIp) Want to discuss these findings and learn more about Binary Quantization? [Join our Discord community.](https://discord.gg/qdrant)
Learn more about how to boost your vector search speed and accuracy while reducing costs: [Binary Quantization.](https://qdrant.tech/documentation/guides/quantization/?selector=aHRtbCA%2BIGJvZHkgPiBkaXY6bnRoLW9mLXR5cGUoMSkgPiBzZWN0aW9uID4gZGl2ID4gZGl2ID4gZGl2Om50aC1vZi10eXBlKDIpID4gYXJ0aWNsZSA%2BIGgyOm50aC1vZi10eXBlKDIp)
@@ -0,0 +1 @@
{}
@@ -0,0 +1 @@
{"state":{"hard_state":{"term":0,"vote":0,"commit":0},"conf_state":{"voters":[3820286201499313],"learners":[],"voters_outgoing":[],"learners_next":[],"auto_leave":false}},"latest_snapshot_meta":{"term":0,"index":0},"apply_progress_queue":null,"peer_address_by_id":{},"this_peer_id":3820286201499313}
Binary file not shown.

After

Width:  |  Height:  |  Size: 126 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 196 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 174 KiB

@@ -0,0 +1,56 @@
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<svg
id="Layer_1"
data-name="Layer 1"
viewBox="0 0 512 512"
version="1.1"
sodipodi:docname="icon.svg"
inkscape:version="1.2.2 (b0a8486541, 2022-12-01)"
xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
xmlns="http://www.w3.org/2000/svg"
xmlns:svg="http://www.w3.org/2000/svg"
xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
xmlns:cc="http://creativecommons.org/ns#"
xmlns:dc="http://purl.org/dc/elements/1.1/">
<sodipodi:namedview
id="namedview11"
pagecolor="#ffffff"
bordercolor="#666666"
borderopacity="1.0"
inkscape:showpageshadow="2"
inkscape:pageopacity="0.0"
inkscape:pagecheckerboard="true"
inkscape:deskcolor="#d1d1d1"
showgrid="false"
inkscape:zoom="1.7011719"
inkscape:cx="166.64983"
inkscape:cy="256"
inkscape:window-width="2560"
inkscape:window-height="1048"
inkscape:window-x="0"
inkscape:window-y="32"
inkscape:window-maximized="1"
inkscape:current-layer="Layer_1" />
<defs
id="defs4">
<style
id="style2">.cls-1{fill:#141f38;}</style>
</defs>
<title
id="title6">atom-glyph</title>
<path
class="cls-1"
d="M332.89,256a64.11,64.11,0,1,1-64.11-64A64.05,64.05,0,0,1,332.89,256ZM503.35,391c-15.42,26.7-49.51,39.7-93.81,39.7A263.61,263.61,0,0,1,379,428.82a38.39,38.39,0,0,1-50.35,53.56C310.78,501.26,290.41,512,268.78,512c-37.94,0-72-32.93-95.49-85.23a279.84,279.84,0,0,1-45.25,4c-44.3,0-78.39-13-93.81-39.7-19.94-34.53-4.38-84.58,36.21-135a283.53,283.53,0,0,1-19.7-27.62,38.38,38.38,0,0,1-25.09-72.54c-.13-12.75,2.59-24.53,8.57-34.89,15.42-26.7,49.51-39.7,93.81-39.7a279.86,279.86,0,0,1,45.27,4C196.77,32.92,230.85,0,268.78,0s72,32.93,95.49,85.22a295.31,295.31,0,0,1,33.58-3.66A38.44,38.44,0,0,1,473.65,94c12.91,6.47,23,15.42,29.7,26.93,19.94,34.52,4.37,84.58-36.21,135C507.72,306.46,523.3,356.52,503.35,391ZM373.94,109.61a403.67,403.67,0,0,1,17.91,74.59A459.57,459.57,0,0,1,450,236.31c3.09-4,6.08-8,8.85-12,25.87-37.31,34-70.33,22.28-90.6-4.28-7.42-11-12.82-18.72-16.76a38.45,38.45,0,0,1-61.18-9.92A260,260,0,0,0,373.94,109.61Zm-6,204.07A481.44,481.44,0,0,0,371.35,256a481.52,481.52,0,0,0-3.42-57.69,546.29,546.29,0,0,0-99.15-57,546.33,546.33,0,0,0-99.15,57.05,488.28,488.28,0,0,0,0,115.36,542.85,542.85,0,0,0,47.91,31,542.51,542.51,0,0,0,51.25,26,542.37,542.37,0,0,0,51.25-26A542.77,542.77,0,0,0,367.94,313.67Zm-35.09,53.21c-10.51,6-21,11.63-31.54,16.83a408.46,408.46,0,0,0,47.25,13.91,349.6,349.6,0,0,0,13.72-48.92C352.79,354.94,343,361,332.84,366.88Zm-128.11,0c-10.16-5.85-19.95-11.94-29.42-18.18A349.52,349.52,0,0,0,189,397.62a408.5,408.5,0,0,0,47.25-13.91C225.77,378.51,215.24,372.93,204.73,366.88ZM140.57,256c0-12.22.46-24.22,1.29-36a404,404,0,0,0-37.69,36,404,404,0,0,0,37.69,36C141,280.21,140.57,268.21,140.57,256Zm64.16-110.89c10.51-6,21-11.63,31.55-16.84A408.48,408.48,0,0,0,189,114.36a349.61,349.61,0,0,0-13.72,48.92C184.78,157,194.57,151,204.73,145.11Zm128.11,0C343,151,352.79,157,362.26,163.28a349.58,349.58,0,0,0-13.72-48.92,408.72,408.72,0,0,0-47.24,13.91C311.81,133.48,322.34,139.06,332.84,145.11ZM397,256c0,12.21-.46,24.21-1.29,36a404,404,0,0,0,37.7-36,403.8,403.8,0,0,0-37.7-36C396.54,231.78,397,243.78,397,256ZM199.05,90.44a464.64,464.64,0,0,1,69.73,23,464.72,464.72,0,0,1,69.73-23c-19.61-41-45.37-64.84-69.73-64.84S218.66,49.48,199.05,90.44ZM87.57,236.31a459.65,459.65,0,0,1,58.15-52.11,403.6,403.6,0,0,1,17.91-74.59,250,250,0,0,0-35.6-2.76c-20.94,0-58.08,3.49-71.6,26.89-3.6,6.23-5.31,13.68-5.22,22.07A38.27,38.27,0,0,1,71,212.42c2.38,3.92,4.92,7.89,7.72,11.92S84.48,232.33,87.57,236.31Zm76.06,166.07a403.65,403.65,0,0,1-17.91-74.59,459.74,459.74,0,0,1-58.15-52.11c-3.09,4-6.08,8-8.85,12C52.85,325,44.73,358,56.44,378.25,70,401.64,107.1,405.14,128,405.14A249.86,249.86,0,0,0,163.63,402.38ZM310.76,464a38.19,38.19,0,0,1,11.84-46.61,479.66,479.66,0,0,1-53.81-18.83,464.68,464.68,0,0,1-69.74,23c19.61,41,45.37,64.85,69.74,64.85C282.8,486.4,297.26,478.47,310.76,464Zm148.1-176.32c-2.77-4-5.76-8-8.85-12a459.54,459.54,0,0,1-58.16,52.11,403.67,403.67,0,0,1-17.91,74.59,250,250,0,0,0,35.6,2.76c20.94,0,58.08-3.49,71.6-26.89C492.85,358,484.73,325,458.86,287.65Z"
id="path8"
style="fill:#ffffff" />
<metadata
id="metadata296">
<rdf:RDF>
<cc:Work
rdf:about="">
<dc:title>atom-glyph</dc:title>
</cc:Work>
</rdf:RDF>
</metadata>
</svg>

After

Width:  |  Height:  |  Size: 4.4 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 MiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 MiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 834 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 834 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 MiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.1 MiB