[Feature Request]: llama-index can't reduce the dim of embedding vector, need to add option to control num of vec features

run-llama / llama_index

LlamaIndex is a data framework for your LLM applications

https://docs.llamaindex.ai

MIT License

33.72k stars 4.74k forks source link

[Feature Request]: llama-index can't reduce the dim of embedding vector, need to add option to control num of vec features #14394

Open zydmtaichi opened 3 weeks ago

zydmtaichi commented 3 weeks ago

Feature Description

I initial a vector store by MilvusVectorStore and set the dim as 128. my program use the embedding model milkey/gte:large-zh-f16 through OllamaEmbedding which returns 1024 features in embed vector. it raised a error when the dim between store and embedding mechanism is not equal. I want to know if there is some method to apply option to change the dim of embedding vector.

Reason

No response

Value of Feature

No response

logan-markewich commented 3 weeks ago

You need to recreate the vector store from scratch, it can't be changed on the fly