Emmanuel Katto Dubai : Exploring Alternatives to Custom Vectorizer for Weaviate Cloud

HI Everyone, I’m Emmanuel Katto from Dubai, United Arab Emirates (UAE) We’re currently using PyTorch and the Transformers library in a custom Vectorizer class with the model sentence-transformers/all-MiniLM-L6-v2 to vectorize SQL, Python, and YAML files before sending them to Weaviate Cloud. However, this setup relies on a resource-intensive PyTorch and Transformers stack.

Has anyone successfully set up a similar process directly within Weaviate Cloud, or found a more lightweight alternative for vectorization? Any suggestions for optimizing or simplifying this workflow would be greatly appreciated!

Regards
Emmanuel Katto

hi @emmanuelkatto24 !!

We will soon allow running this kind of models from within Weaviate Cloud.

You can ask to be notified about the availability of this new feature here:

Meanwhile, one thing you could do is running your inference model in your own cloud, provide an openai like api, and point your collection there.

For example, check this doc on Weaviate + KubeAi:

Let me know if that helps!