Amazing work! Is it possible to finetune this model on your own code, or a subset? Finetuning this model on pytorch code to help with tensor manipulation would be awesome!