Python + AI: Embedding Vectorales
In our second session of the Python + AI series, we'll delve into a different type of model: vector embedding.
A vector embed is a way to encode text or images as an array of decimal numbers. This makes it possible to perform similarity searches across various types of content.
In this session, we'll explore different vector embedding models, such as OpenAI's text-embedding-3 series, using both visualizations and Python code. We'll compare distance metrics, use quantization to reduce the size of the vectors, and test multimodal embedding models.
If you want to follow the live examples, make sure you have a GitHub account.
Check out the Learn collection: https://aka.ms/PythonAI-Learn/reactor/yt
Explore the episode slides and resources: https://aka.ms/pythonia/recursos/yt
Continue the discussion on Discord: https://aka.ms/pythonia/discord
Develop AI applications with Python: https://aka.ms/eHub/PythonAI
Open the code on GitHub Codespaces: https://aka.ms/python-openai-demos
Chapters:
00:06 - Welcome and code of conduct
01:39 - Global participation and community on Discord
02:01 - Introduction to the Python + AI series
04:05 - What are vector embeddings and why do they matter? 08:02 - Practical Applications of Embeddings
09:59 - How Embeddings Are Generated
19:12 - Similarity Measurement: Cosine Similarity
22:08 - Practical Example: Similarity Between Words
25:00 - Vector Search: Exhaustive vs. Approximate
34:05 - ANN Algorithms and Vector Databases
40:01 - Optimization Techniques: Quantization and Dimensionality Reduction
47:43 - Real Impact on Storage and Performance
58:42 - Next Steps and RAG Session
#microsoftreactor #learnconnectbuild
[eventID:26281]
In questa pagina del sito puoi guardare il video online Python + AI: Embedding Vectorales della durata di ore minuti seconda in buona qualità , che l'utente ha caricato Microsoft Reactor 01 gennaio 1970, condividi il link con amici e conoscenti, su youtube questo video è già stato visto 11,996 volte e gli è piaciuto 990 spettatori. Buona visione!
