Train llama with lora on one 4090 and merge weight of lora to work as stanford alpaca.
$ claude mcp add alpaca-weight \ -- python -m otcore.mcp_server <graph>
⬇ download graph artifact