I deployed a recommendation model. Testing Models In Production using Interleaving Experiments.

  Рет қаралды 1,675

Underfitted

Underfitted

Күн бұрын

Пікірлер: 8
@dimitriskapetanios294
@dimitriskapetanios294 8 ай бұрын
Very insightful, thanks Santiago!
@lokeshsharma4177
@lokeshsharma4177 8 ай бұрын
YAA - You Are Awesome 🙏🙏🙏🙏
@BiXmaTube
@BiXmaTube 8 ай бұрын
Sorry for the off topic, Really need proper pdf parsing ai that I can run on a cloud server without gpu. Extracting text, tables and images and arranging it in a db based on a prompt that puts each data in the right table. For example store address from the pdf in an address field in the db, images links in their own field as the images are stored in the pdf created folder, etc...That will be amazing if you can find something like that.
@X9212
@X9212 8 ай бұрын
How do you deal in the event of same recommendations e.g. 1 reco of each model is the exact same. ? Not take into account those events?
@brenswen_
@brenswen_ 8 ай бұрын
Interleaving has always been an interesting idea to me, but at the end of the day you're not measuring the experience that the user will see if you roll out the new model to 100% like you would if you did a typical A/B test. In my experience, you can get signal from users in the first day or two if you're tanking the experience, and if you have guardrail metrics, you can turn the test off
@underfitted
@underfitted 8 ай бұрын
Right. Goal is to evaluate how good are the recommendations from the candidate model. If people aren’t choosing those, turn it off.
@Jasonmayers12
@Jasonmayers12 8 ай бұрын
Sir, I know this may be irrelevant related to this video, but i am actually following you since 2021 and your content is really amazing. But can you make a tutorial on this youtube video that i found kzbin.info/www/bejne/gJfVqHt4ZqmZbLMsi=Hn-H-zJ4asfNI15y I can't find any tutorials being made on this and it's very underrated in my opinion determined by its applications in various fields
A gentle introduction to RAG (using open-source models)
50:10
Underfitted
Рет қаралды 14 М.
Why are vector databases so FAST?
44:59
Underfitted
Рет қаралды 20 М.
So Cute 🥰 who is better?
00:15
dednahype
Рет қаралды 19 МЛН
Support each other🤝
00:31
ISSEI / いっせい
Рет қаралды 81 МЛН
It’s all not real
00:15
V.A. show / Магика
Рет қаралды 20 МЛН
When you have a very capricious child 😂😘👍
00:16
Like Asiya
Рет қаралды 18 МЛН
Visualizing transformers and attention | Talk for TNG Big Tech Day '24
57:45
How I made $600,000 freelancing on Upwork.
48:16
Underfitted
Рет қаралды 12 М.
[Webinar] How to Build a Modern Agentic System
1:00:55
Arthur
Рет қаралды 10 М.
AI can't cross this line and we don't know why.
24:07
Welch Labs
Рет қаралды 1,5 МЛН
Transformers (how LLMs work) explained visually | DL5
27:14
3Blue1Brown
Рет қаралды 4,3 МЛН
The Elegant Math Behind Machine Learning
1:53:12
Machine Learning Street Talk
Рет қаралды 159 М.
How to train a model to generate image embeddings from scratch
51:44
The $1,000,000 problem AI can't solve
13:24
Underfitted
Рет қаралды 2,2 М.
How might LLMs store facts | DL7
22:43
3Blue1Brown
Рет қаралды 975 М.
So Cute 🥰 who is better?
00:15
dednahype
Рет қаралды 19 МЛН