Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval (#1934)

* Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval

* Mark Aamir as guest

Co-authored-by: Pedro Cuenca <[email protected]>

* Update memory/disk usage with 41M embeddings

* #Demo -> #demo

* Update Aamir's username

* Apply suggestions from review

Thanks a ton, Pedro!

Co-authored-by: Pedro Cuenca <[email protected]>

* Replace most footnotes with links

* Remove all comments

* add comment about 92.5% without rescoring in binary quant.

* Separate the code blocks for binary quantization

* Apply suggestions from review

Co-authored-by: Pedro Cuenca <[email protected]>

* Fix incomplete links

* Add clarification about the difference between scalar and binary

* Apply suggestions from review

Thanks Omar!

Co-authored-by: Omar Sanseviero <[email protected]>

* Mention that embedding quantization is not like model quantization

---------

Co-authored-by: Pedro Cuenca <[email protected]>
Co-authored-by: Omar Sanseviero <[email protected]>

Author: 3 people

_blog.yml

@@ -3707,4 +3707,17 @@
   tags:
     - leaderboard
     - arena
-    - collaboration
+    - collaboration
+
+- local: embedding-quantization
+  title: "Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval"
+  author: aamirshakir
+  guest: true
+  thumbnail: /blog/assets/embedding-quantization/thumbnail.png
+  date: Mar 22, 2024
+  tags:
+    - nlp
+    - community
+    - guide
+    - collaboration
+    - research