Would be cool to see the performance difference for llama2.c or see it work for gddr on gpus too with nanogpt, though I guess the latter might or might not be p

Would be cool to see the performance difference for llama2.c or see it work for gddr on gpus too with nanogpt, though I guess the latter might or might not be possible because of architecture differences.