Ability to Set Dimensions in the Embedding Request #7602
-
Right now there is no way to pass dimensions as a part of the GenerateEmbeddingsAsync method in Semantic Kernel because the signature does not allow it. The only way seems to be to define the dimension at the time of adding the service to kernel. We would definitely like to have the ability to set the dimension as part of the individual request as a part of executionsettings similar to the pattern followed in the TextGeneration and Chat signatures. Is this functionality planned as a part of the embedding re design? Also is there an interim way to accomplish this without having to write a custom connector? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
Hi @jmurkothintel, thanks for your comment/question. We're working on new memory abstractions which allows for a much more custom approach to dealing with memory and embeddings. Are you targeting Python or dotnet (my guess is dotnet based on the method name you mentioned above)? For Python you can view it here: https://github.com/microsoft/semantic-kernel/blob/feature-python-memory/python/samples/concepts/memory/new_memory.py For dotnet you can view it here: https://github.com/microsoft/semantic-kernel/blob/feature-memory-service/dotnet/samples/Concepts/Memory/VectorStore_DataIngestion_Simple.cs#L82 |
Beta Was this translation helpful? Give feedback.
Hi @jmurkothintel, thanks for your comment/question. We're working on new memory abstractions which allows for a much more custom approach to dealing with memory and embeddings. Are you targeting Python or dotnet (my guess is dotnet based on the method name you mentioned above)?
For Python you can view it here: https://github.com/microsoft/semantic-kernel/blob/feature-python-memory/python/samples/concepts/memory/new_memory.py
For dotnet you can view it here: https://github.com/microsoft/semantic-kernel/blob/feature-memory-service/dotnet/samples/Concepts/Memory/VectorStore_DataIngestion_Simple.cs#L82