Get the multiple layer attention map from ESMC

[Luchixiang]

Thanks for your wonderful work. I'm wondering is it possible to get the multiple layer attention maps from ESMC-6B via ESM3ForgeInferenceClient?

[lhallee]

My group wrote an implementation of ESMC called ESM++. We support attention maps when output_attentions are passed. The 300 and 600 million parameter models are currently supported - thought it may be helpful to you.