This may happen if you chose a model which is larger then your available unified memory.
Which version of llama2 did you choose? And how much unified memory do you have?
Which version of llama2 did you choose? And how much unified memory do you have?