I also have the same issue . I have been working with the same model and system . Most of the time my kernal will die or the system will hang for longtime . I believe it because of the mac spec and its Gpu Intel UHD Graphics 630 1536 MB is not sufficient for many llama models . To address this I have few suggestions .

1. Use Quantized llama models . These are lightweight models .
2. if possible try to get an EGPU for your mac .
3. upgrade to PC with better specifications .