Embedding Function

#11
by dandre0102 - opened

Hey there,

I've been trying to find and make use of the embedding function LLava uses - but so far no luck.
Any pointers on how to proceed would be much appreciated.

Thanks in advance

Did you figure it out? I was thinking of using the outputs of vision and language models separately to create the embeddings (using the last hidden state), but I don't know if this is correct + also if there's a way to apply a projection to the outputs.

Sign up or log in to comment