Multi-modal data2vec (joint training)

[netw0rkf10w]

❓ Questions and Help

Hello,

I see that you have implemented multi-modal data2vec (in the sense of jointly training the modalities), but the paper did not mention this. I was wondering if you have actually tried this and obtained some findings (maybe it didn't work?). Could you please kindly share some information on this? @alexeib @michaelauli

Thank you very much in advance!