Experience the power of our new embedding models, trained on legally approved data for business usage
Exciting news! Our latest release introduces the watsonx Embeddings API, allowing you to generate vectorized representations of input text and capture semantic relationships
𝗪𝗵𝗮𝘁'𝘀 𝗻𝗲𝘄:
1. Embeddings generation endpoint in our REST API
2. Two initial models: slate.125m.english.rtrvr and slate.30m.english.rtrvr (knowledge distilled)
3. Accessible through REST API endpoint or Python SDK
𝗪𝗵𝗮𝘁 𝗮𝗿𝗲 𝘁𝗲𝘅𝘁 𝗲𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀?
Text embeddings are numerical representations of sentences or passages, allowing computers to quickly and efficiently compare and analyze semantic meanings.
𝗪𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻𝘀 𝗳𝗼𝗿 𝘆𝗼𝘂:
Embedding models are crucial for Generative AI applications, offering semantic and syntactic fidelity with efficient computing and storage.
𝗦𝘁𝗼𝗿𝗲 𝗮𝗻𝗱 𝗨𝘀𝗲 𝗘𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀
Store vectors in a vector database like Milvus and use them to compare passages in real-time or reference stored vectorized text passages.
𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 (𝗥𝗔𝗚)
Use embeddings to enhance AI model accuracy and reliability with facts from external sources.
𝗔𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲𝘀 𝗼𝘂𝗿 𝗦𝗹𝗮𝘁𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝗼𝗳𝗳𝗲𝗿 𝗼𝘃𝗲𝗿 𝗼𝘁𝗵𝗲𝗿 𝟯𝗿𝗱-𝗽𝗮𝗿𝘁𝘆 𝗲𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹𝘀:
1. Performance matching or exceeding market leaders in retrieval benchmarks.
2. Exclusively trained on legally approved and commercially viable data for enterprise usage.
3. Scalable, fully-managed, and integrated with the broader watsonx portfolio.
![diagram](https://media.licdn.com/dms/image/D5622AQGf6WZflcmoZA/feedshare-shrink_2048_1536/0/1713792634727?e=1716422400&v=beta&t=AFLcrA8DrjO-QxW9MC0S9riwTPJ30MIfYYtIRnbAus8)
𝗚𝗲𝘁 𝗦𝘁𝗮𝗿𝘁𝗲𝗱 𝗻𝗼𝘄 𝘄𝗶𝘁𝗵 𝘁𝗵𝗶𝘀 𝗥𝗔𝗚 𝗧𝘂𝘁𝗼𝗿𝗶𝗮𝗹!
Here's a Python Notebook leveraging our watsonx Granite Model Series, Chroma, and LangChain to answer questions: https://lnkd.in/gZB-qk-f
#watsonx.ai#GenerativeAI------------------------------
Armand Ruiz
------------------------------