The easiest way to run alpaca Lora locally is with this little known fork [1] that uses Docker. You’ll be up and running in about 20 min with pretty much any modern consumer Nvidia GPU.

[1] https://github.com/chris-alexiuk/alpaca-lora