lukewys

/

laion_clap

like 32

Hello,
I would like to use your most recent checkpoints with transformers.ClapAudioModelWithProjection class.
I tried to load the state dict but the keys name are different, and the number of keys as well.

If my understanding is correct, you already did this work for the previous checkpoints (https://huggingface.co/laion/clap-htsat-unfused), so it must be possible. Could you give me some hints on how to proceed? Thank you!