Building an Open-source Framework for Generating Embedding Vectors
summary
An open-source framework called towhee provides operator-based pipelines that glue together models, ETL, and vector storage so teams can prototype embedding applications end-to-end. It integrates Hugging Face models and common CV nets like ResNet50 and multimodal models such as CLIP, computes and inserts 1,000 image embeddings into Milvus in about four minutes on a single CPU, and offers ensemble, fine-tuning, Triton/Ray-backed serving and vector DB adapters — the reverse-image search demo runs and returns sensible nearest neighbors.
AI-written summary · about
chapters (8)
- 0:00Introduction and overview
- 1:03Defining unstructured data
- 2:13Understanding embeddings
- 6:34Semantic search applications
- 8:59Challenges in semantic search
- 15:47Introduction to Towhee
- 26:58Code demo and implementation
- 35:30Q&A and closing
related talks
45mA gentle introduction to Vector Databases
50mVector Ops: run vector embedding-powered apps in prod
01h 26mBuilding an Image Search Engine with Generative Computer Vision Models
31mBuild your own search application with Weaviate
50mWhat Comes After ChatGPT? Vector Databases - the Simple & Powerful Future of ML?
28mEmbeddings
57mUsing Vector Databases for Multimodal Embeddings and Search
01h 03mWhat comes after ChatGPT? Vector Databases - the Simple and powerful future of ML?