I could not find any documentation that says v68 supports that. This features is crucial for any transformer-based model.
Dear customer,we are currently syncing this issue internally.
Dear customer
Qualcomm’s model conversion tool indicates that it does not support dynamic tensor dimensions.