Adding `safetensors` variant of this model

Add info about possible retraining requirement for trigger happy model

Some more details

provide minimal description

upload all

initial commit