How was expored from m3

#1
by bergum - opened

I'm wondering if you could share how the model was exported to onnx? I see that it has two outputtoken_embeddings and sentence_embedding

Owner

Still a WIP, but it's with this PR.

Hi @bergum I just wrote code leveraging HF Optimum library to export BGE-M3 to single ONNX model which outputs dense, sparse and colbert vecs all at once. Let me know if you are interested in it

That would be awesome @aapot !

@bergum I just published my ONNX model and also needed scripts to export the ONNX model here: https://huggingface.co/aapot/bge-m3-onnx
Let me know if you have any questions!

Wonderful @aapot !

Sign up or log in to comment