How to parameter efficient finetune Decoder in encoder-decoder model?

have you tried fixing it? i faced the same exact problem