Embedding Extraction

Open tashjeans opened this issue 9 months ago • 1 comments

Hi Boltz community, I have successfully installed and extracted a prediction as a test case, but I was wondering if there were internal embeddings in the model that could be extracted for downstream use? If so, are these similar to AF or ESM embeddings, where there are three sets of node/edge and MSA embeddings? And is there any information on where exactly in the model these are found/how to extract them! Thanks :)

Jun 02 '25 10:06 tashjeans

I was looking for an answer to the same question, and apparently it has been answered before!

https://github.com/jwohlwend/boltz/issues/87

Kind regards Dimi

Jun 04 '25 14:06 dimi-swiftpharma